Jack Crossfire's blog View Details
Posted by Jack Crossfire | Mar 14, 2020 @ 10:51 PM | 7,869 Views
What a lion actually looks like with his dog.
Posted by Jack Crossfire | Mar 14, 2020 @ 10:48 PM | 7,838 Views
Was hoping it just cracked acrylic, but the lunchbox cracked too. Tried separating threadlock from plastic with a washer, but the threadlock most have gone right through. Just a matter of applying threadlock after tightening the nut.


There should be enough friction from the washer to keep it together, but if it ever has to be taken apart, it's going to need a reinforcing layer of teflon. The chassis is extremely rare on ebay or a $130 tag, in today's money.
Posted by Jack Crossfire | Feb 09, 2020 @ 12:02 AM | 4,753 Views
It didn't take long for the lion kingdom to realize it was running hundreds of miles in silence. The Anker was no louder than a phone speaker. There are 2 ways to make a bluetooth speaker louder. They all currently have a separate radio & amplifier. The HT8692 amplifier in the Anker has a gain adjustment resistor, but it has a very low maximum gain. Anything below 65k made no difference.
The next step was putting an op-amp between the ATS2815 radio & amplifier. All it took was a bare LF353 with a gain pot. The Anker provides a virtual ground & dividing resistors to combine the 2 channels.

The maximum gain of this circuit was limited by the input offset voltage of the op-amp. Above the maximum gain, the DC offset of the output mutes the signal. The lion kingdom tuned it just below the maximum gain & relied on the phone volume to limit distortion. It's amazing how much power these speakers have put out, over the last 5 years, without blowing up.
Posted by Jack Crossfire | Jan 16, 2020 @ 12:40 AM | 6,767 Views
After literally years of looking for a compact way to power a bluetooth speaker from the mane battery, the reason why they never worked was finally found. It wasn't the large filter capacitor or the ripple but the $1 car chargers only handling 1 Amp. When they were upgraded to 2A, the speakers started working with no changes. 1A is not enough to power a speaker, despite all the current measurements coming in below 1A.

The easiest solution was always to not hack anything but stick an original speaker in with a plane old USB cable & car charger. The hacking apart of car chargers & speakers was to make them more compact & lighter. Now the convenience of only having to charge the mane battery & transmitter battery, with no beeping after 10 miles, is finally known.
Posted by Jack Crossfire | Jan 09, 2020 @ 01:13 AM | 3,387 Views
The answer is no. You can't power a bluetooth speaker from a $1 car charger. The $1 car chargers are only useful for charging because they make too much ripple. Only a very large filter capacitor can reduce the ripple enough, in which case a battery is smaller. A giant filter capacitor is still more convenient than a battery. After years of taking 3 batteries out of the vehicle for charging, it's so convenient to only have to worry about 2.

The car chargers can be tested with an 8ohm load. The ripple is affected by output capacitance & battery impedance. Newer batteries can power bluetooth speakers more easily. The Rudung creates 100mV of ripple. The 10000uF output capacitor + 1R resistor reduces it to 50mV. The 10000uF with a new battery produces no ripple. An inductor on the output makes no difference.
Posted by Jack Crossfire | Jan 07, 2020 @ 12:48 AM | 5,493 Views
The $1 buck converters now go up to 2A, fixed voltage of course. Almost worth building a dummy load to see how long they really handle 2A before blowing up.
Posted by Jack Crossfire | Jan 04, 2020 @ 04:21 AM | 7,529 Views
$1 car USB chargers were a good substitute for expensive voltage converters for many years. Just replace the feedback resistors with a pot & they could make any voltage at 1A.
Of course, if you didn't already have a bag of pots & caps, they would be a lot more than $1.

They were useless in their intended role as car chargers because the caps exploded in the heat.

The party ended when the Chinese finally moved to fixed voltage regulators. Any application would need to run on 5V now.
Posted by Jack Crossfire | Jan 03, 2020 @ 12:28 AM | 4,892 Views
Don't plug a 4.2V speaker into 12V or it'll let out the magic smoke. Fortunately, it only blew up the amplifier chip, a HT6871. The radio & leds were still intact. It would take 1 month to order a new amplifier chip from China, so went ahead & got a new speaker. With this experience, it was decided to put in different connectors for 12V than 4.2V.

It was such a good speaker during its 3 year lifespan, the lion kingdom put in a different amplifier chip & revived it. There were some amplifier chips from a terrible speaker with the same pinout. The new speaker arrived anyway & was a better deal.

https://www.amazon.com/gp/product/B01HTH3C8S

Manely, the newest speaker turns on right when power is applied instead of requiring holding a button after applying power. It's all a matter of luck, but money buys a lot more online with the 1 week delay than locally.

The sub $25 can speakers have been the right size for robots.


The new amplifier chip uses a boost converter to make 7V. It would be simpler just to use a speaker with higher impedance or a battery with higher voltage to get the same power, but apparently it's cheaper to use higher voltage with standard speakers & a battery voltage which doesn't require a boost converter for charging.
Posted by Jack Crossfire | Dec 18, 2019 @ 12:33 AM | 9,818 Views
The decorations keep getting better, with the discovery of brown pipe cleaners at Hobby Lobby & the order of giant red LEDs for the LED animation.
Posted by Jack Crossfire | Oct 22, 2019 @ 02:33 AM | 3,326 Views
A scooter ran a red light at 25mph & smashed straight into it. Miraculously, only the wheel was smashed. He broadsided it & dragged it a few feet, but manely took out the food it was carrying. Scooters & bikes are in the grey area of not having red light rules enforced & not having speed limits.

In regular use, the lion kingdom has had 1 wreck every 2 years & all in crosswalks or sidewalks. The 1st wreck was an SUV turning right, destroying the cargo area, damaging the chassis & 1 wheel.

Ground vehicles which coexist in the road network have similar expenses as flying, but instead of impacts with the ground, it's impacts with full sized vehicles. Fortunately, they don't damage the other vehicles or cause injuries, but when cars impact robots, car insurance doesn't cover it & it's pretty much the robot owner's expense.

They're quick with the sorry's & the excuses but no driver is willing to pay for it. Lions don't pressure anyone for money because of what happened to quad copters. If there's any hint of liability, they'll regulate the ground robots into the stone ages. They're already banned in almost as many areas as quad copters.

The mane problem is they're harder to see than humans. Near misses with cars happen every month, usually when cars make right turns. High acceleration & planning for it is key. The lion kingdom usually has the robot drive ahead, in order to bait cars. If the car keeps going, there's a good chance they don't see the lion.

There were a few right turns that would have impacted lion instead of robot. The robots are all intended to be expendable, but there's still a desire to prevent one from becoming a total $500 loss, in today's money.
Posted by Jack Crossfire | Oct 20, 2019 @ 11:32 PM | 4,964 Views
The lion kingdom can remember no time when it needed 2 power supplies simultaneously. They were only used for remembering different settings or having different connectors. Manely, they were hoping for a future need that never came.


A Rudung was sacrificed. The new butter surprise created a stable 9V from a 12V battery. It had a 0.5V dropout.

Subjectively, the results were more realistic than without a regulated voltage. There was a steep drop in PWM for the 1st turns, then a leveling off. Finer precision was still a matter of complete drives & measuring charge.

The alignment affected downhill drives more than uphill. It might have been because more weight was on the front wheels when going downhill.


The 1st 9.2 mile drive burned 246mAh/mile.
Posted by Jack Crossfire | Oct 19, 2019 @ 06:41 PM | 4,773 Views
It broke after only 6 months. Knocking the wheels the right way, rather than old age, is how they break. The only replacement part is the fabled Tamiya Shaft Bag 58347, long out of production. There are aluminum ones in China with a 6 week waiting time.



Lacking any replacement parts, the lion kingdom attempted a teflon repair. The bolt isn't long enough to reach through the teflon. Teflon is softer than the original nylon.


During this process, it became clear the steering rods could be 1 hole shorter, so the wheels would have to be realigned for the teflon repair & aligned again next year when the China shipment arrived. All this wheel alignment had the lion kingdom searching for a better way. Lions traditionally aligned the wheels by driving a mile, recharging the battery, & measuring the charge. The charge depended on the wheel alignment as well as the battery temperature, & how charged it was yesterday.


The ideal way is to drive a certain distance & record the PWM, but this requires a constant battery voltage. A Rudeng/RIDEN voltage regulator is still the ideal solution for getting a portable, constant voltage. Getting one is another 6 week China shipment.

It was while searching for a voltage regulator that lion kingdom realized this was what Sparkfun sold, 15 years ago. What did Sparkfun sell nowadays? Manely dumbed down educational kits for large schools, big ticket items for corporations, & some standard connectors. Their power supply offerings where now the laptop bricks office supply stores sold 20 years ago instead of the bare boards they used to sell.


Exotic parts for starving college students & programmers are now only available in China. The next option is building a fixed voltage buck converter out of digikey parts, the same way lions would have done it 20 years ago.

There may just be no money to be made in exotic parts just for building other things. They'd rather buy something prebuilt.
Posted by Jack Crossfire | Oct 02, 2019 @ 02:31 AM | 2,909 Views
Very disappointing watching the BFR becoming more & more complicated over the years. The original concept was the simplest because it had 1 purpose: moving humans & cargo to Mars. For every trip, it landed with everything it took off with. They could load it with Martian regolith for the return trips. This way, the center of gravity was the same for every trip. The nose payload would always balance the weight of the engines, allowing it to reenter as a blunt body. Spaceships need to reenter as a blunt body rather than nose first, to lighten their heat shields.

It had header tanks on the top of the mane tanks. The header tanks fueled the landings. By having them as high as possible, it would reduce the amount of ballast required to weigh the nose down during reentry. By having them inside the mane tanks, they would be insulated for long duration flights.

The problem was to make money, the vehicle had to be repurposed for the core satellite business. Payloads would be dropped off in Earth orbit & the vehicle had to return without a payload. This was the same problem NASA encountered with the shuttle.


Like NASA, SpaceX added wings to try to make it pitch down without a payload. The problem was to still satisfy its original role of moving humans to Mars, the wings needed variable lift. To give the wings variable lift, they had to move forward rather than reenter as a blunt body. This required a heavier heat shield.

The shuttle worked around the...Continue Reading
Posted by Jack Crossfire | Sep 17, 2019 @ 07:14 PM | 3,784 Views
Quite a shock when the fixture of Gootube recommendations for the last year, the one of a kind PZL-104 modified with a turbine & extra large landing gear was destroyed. He took off in a strong crosswind & got blown over. This is why there are maximum windspeeds for aircraft.



He provided 10 minutes of video of him beating himself up but showed very little of the wreck. His passengers miraculously survived without any injury. It must have been a very forgiving airplane for him to be so lethargic about the weather.


In another time, it probably could have been rebuilt with parts from another PZL-104 & another engine. He probably doesn't have enough money to do the job. He seemed to document at least some of the build process, so that's more than the Smithsonian had for rebuilding the Wright flier.
Posted by Jack Crossfire | Jul 04, 2019 @ 03:17 PM | 5,464 Views
Stabilized gopro 7:

Gopro 7 60fps footage (13 min 50 sec)


Gopro 7 stabilization comparison (5 min 37 sec)
...Continue Reading
Posted by Jack Crossfire | Jul 01, 2019 @ 02:11 PM | 6,170 Views
The price of the labor means nothing if they're good. Maybe Hindustan Computers Limited isn't the best at writing airplane software. Some of the smartest people in the world are Indians making $9/hour. Some of the dumbest are USicans making $400/hour. If wealth was fairly distributed, there would be a recession.

Software written to buried sections within a million page spec tends to miss the details that come from casual use. The relevance of a piece in the big picture gets lost. What happens if the piece fails gets lost.

Boeing has a long feedback loop beginning with a guy in Seattle who just writes a spec, a guy in Mumbai who just implements it, & a pilot who just tests it years later. We've all applied to large aerospace companies. All they want are PhD's in exactly 1 function, who have spent their entire lives doing 1 function without ever seeing anything else.

Maybe if the guy in Seattle implemented it & had a way to test it instead of writing out a spec for someone else, the loop would be shorter.
Posted by Jack Crossfire | May 01, 2019 @ 12:59 AM | 4,108 Views
Ironically, an average game written for the Commodore 64 today would actually be more popular than a mobile app, just because a lot less games are written for it & because there's a lot of interest in retro computing. Wanted to make a Robotech game for the C64.


While the failure was attributed to lack of focus & development tools, what the lion kingdom wanted was really too ambitious for the hardware. A full recreation of the TV universe, with a fully functional veritech fighter that could fly anywhere in the TV show was not possible. Lions even wanted the player to drop into a different screen in battleloid mode, like the TV show.


What was possible was a recreation of single locations in the TV universe, with the veritech in just 1 of its 3 modes at a time, & a single opponent. A complete level could be the 1st encounter with an alien in battleloid mode. The emphasis could be replicating the visuals & the animation style of the TV show, but not gameplay.


Today, there are much better Robotech games, but none recreate what really made the show work: the love story & the animation style.
Posted by Jack Crossfire | Apr 23, 2019 @ 11:30 AM | 3,905 Views
The Tesla autonomy show had a lot of nuggets about what the typical self driving car startup does. There's a chip which is a scaled up version of a GPU, using stock IP that doesn't use very current feature sizes, but marketed like crazy as a revolution. Something never mentioned before, but which slowly oozed out over the last 5 years is how training the networks still entails manually annotating thousands of images. They have some interpolation, some LIDAR to generate annotations, but most of it is mass Bangalore labor.

A more telling feature was the use of neural networks in place of all conditional statements. Determining if a car was changing lanes was an exercise in thousands of sets of car coordinates manually classified as lane changes. They couldn't just classify it based on x > threshold. Enormous effort was spent doing it with a neural network. How much more effort would have been required to make a pose classifier based on neural networks instead of if statements?


There's still no magic in using a neural network to detect a path. It has to be trained with thousands of images of every section of path in every lighting condition. They all have to be manually annotated or interpolated with path coordinates. All the network does is compress the training set & replay it. The network has to be more like millions of neurons instead of hundreds.
Posted by Jack Crossfire | Apr 02, 2019 @ 12:18 AM | 10,554 Views
It was $400 all in, without the electronicals. After 20 miles, the transmission was extremely noisy, despite lubrication & the sound damper. Despite the noise, the efficiency was a record high of 240mAh/mile empty. 2200kV is a better speed than 2400kV. The hall effect sensor worked without any firmware changes. It actually manetaned the same speed going downhill. There's no brake modulation, so it would have to always deliver the minimum power....Continue Reading