Jack Crossfire's blog View Details
Archive for November, 2017
Posted by Jack Crossfire | Nov 29, 2017 @ 02:01 PM | 3,653 Views
Sadly, the gopro 4 which replaced it was only used once & never produced useful footage. The mane challenges are a vehicle mount which meets the modern height standards, getting shots which the previous cam didn't already get, & the spherecam producing more interesting shots. The only useful part is the microphone. The SD slot wasn't recoverable.

Experiments with the headset & robot cam showed it's worth the investment in a more portable recording solution. The need for VU meters, transport controls, & counters requires a wireless connection to a phone.

It's really hard to justify anything besides a raspberry pi zero W for most of the work. That board only uses 200mA & is the cheapest solution for everything.

A microcontroller is only necessary for converting the ADC's parallel output to an SPI slave. An analog pot is the easiest way to adjust microphone gain. The ADC/microcontroller board would be dedicated just to the headset & interchangeable with a 4 channel version for less portable use. The 4 channel version would have digital pots for line level inputs, another section for monitoring, maybe mic preamps for 2 channels, lots of transistors for a patchbay....Continue Reading
Posted by Jack Crossfire | Nov 28, 2017 @ 11:13 PM | 3,837 Views
Slightly higher again. Replaced the aluminum roll bars with spare CF rods. The most obvious way to mount them was simply hot gluing rivets into the metal. Hot glue rivets have proven very effective at bonding things to metal, but the amount of leverage created by the rods will probably tear them up in a single fall.

Another method involving bolting metal rods to the metal, then strapping the CF to the metal has since emerged. This would be heavier & it's not known exactly how much force would be applied to the hot glue in a fall. The typical falls have been quite benign on the rods.

The CF rods were so much more invisible on video, they showed a lot could be gained from using the least material. The camera had to be rotated to get seamless shots of the lion, which made the rod appear to float more conspicuously in front of the vehicle.
Posted by Jack Crossfire | Nov 26, 2017 @ 10:19 PM | 3,883 Views
Foul weather driving finally seems to be unlocked. Ran through quite a shower without any issues. The only remaneing issue is water getting in the vehicle electronics, but it seems to have made it. Those water levels haven't been reached yet, because the controller was always 1st to go. Further waterproofing of the vehicle is possible, for a severe storm. Had to remove the handle for the 1st time, to fit in the mane's food. Surprised nothing fell out. In hindsight, the handle could have gone back on.
Posted by Jack Crossfire | Nov 25, 2017 @ 06:47 PM | 3,999 Views
With some packing foam fashioned into a windscreen, the Yapster was retired once again. It's still a heavy duty beast with future potential & better sound quality. The next problem is the heft of a phone & monitoring board for recording audio.

The lion kingdom long dreamed of a wireless solution. A small FM transmitter would be able to record the microphone but not send audio from the phone. Despite every effort, there is no easy way to record from a bluetooth headset. The Goog doesn't allow recording from a device which can make phone calls because of laws.
Posted by Jack Crossfire | Nov 24, 2017 @ 10:50 PM | 4,375 Views
The Logitech H111 had a defect in which the microphone was too short. The decision was made to fix it. Despite having some screws, the microphone assembly is not repairable. After cracking plastic welds & boring out the remanes of the microphone tube, it was ready for a longer tube. The original tube was ordinary pipe with 3 steel wires in it. This was replaced with a longer pipe & 2 steel wires because there was no pipe of suitable guage to fit 3 steel wires in.

After hot gluing it back together, it was a mess with much better sound quality. The tube flops around, but is just rigid enough. Unfortunately, the wind screen was lost. It's completely worthless without the wind screen.
Posted by Jack Crossfire | Nov 24, 2017 @ 01:24 AM | 5,290 Views
In what became an epic bodge, the 5 year old headset monitor was finally replaced with a much smaller version. The original design required a large inductor to isolate GSM noise. That was no longer necessary for LTE phones. The original design ganged 2 op amps to drive the speakers in mono. The new design drives the speakers with 2 op amps in stereo.

The original design used cascaded 10x amplifiers in the preamp. The new design uses a single 100x amplifier because 1 op amp had to be bodged for isolating the phone. As in the original design, a voltage follower is required to keep the phone from feeding into the preamp, but the reason wasn't written anywhere. It wasn't to isolate GSM noise.

The original design regulated 12V down to 5V. The new design uses unregulated 4.2V from a battery which lasts 24 hours. The standby current is only 5mA. Peaks get it up to 20mA. It took a lot of bodging to adjust the phone volume. The 5k's double the phone's volume, giving the phone's internal volume control enough room. The only adjustment is the microphone volume.

The decision was made to have the monitor permanently in the headphone line. The headphones were never used for anything else.

There was an idea to use digital mixing in a microcontroller, but this required a lot more parts. The lion kingdom inherited a lot of AD8604's from the day job. These are insane rail to rail 8Mhz op amps.
Posted by Jack Crossfire | Nov 23, 2017 @ 05:15 AM | 4,064 Views
After extensively reviewing every hard drive, it became clear that the entire painstakingly entered DVD database was lost forever. Every time the root hard drive was backed up, the /var & /usr directories somehow got omitted. Mysql had an annoying habit of putting all the databases in /var. The only way to find anything is now to copy every surviving DVD to a 4TB hard drive.

There was 1 more firing up of the 4.3GB Quantum, the lion kingdom's very 1st hard drive. It was apparently reformatted & last used as a dedicated TV, in 2001. The date codes are a bit corrupted. It made strange hisses & clicked a lot, but it still seemed to work after 20 years. DMA had to be configured as 16 bit. Perhaps these were normal sounds, 20 years ago. The lion kingdom has no memory of what this hard drive was last used for, but it does remember greater joy when so much iron was mobilized to access a file.
Posted by Jack Crossfire | Nov 19, 2017 @ 06:42 PM | 2,866 Views
The 1st tests applying anti stiction to spherecam went a lot better, but were not perfect. At 9min/mile, it seemed immensely more stable. At 6min/mile, it was probably about as good as 9min/mile without anti stiction, but not perfect. The P had to be reduced & D increased. Stiction amount had to be increased.

There's no doubt that stiction was the key to stabilizing it, but it was only tested on a 100 meter course. It was also tested with a full 2 traction batteries, spherecam battery, & speaker. It would not be as stable with the minimum payload.

Also, it was discovered that the last 2 guy wires were unnecessary for stabilizing the pole. There was enough static pressure between pole & vehicle to keep it from swaying sideways. A bit of duct tape & shimming would add a little extra friction, even though the adhesive doesn't stick it down. There should be enough stability to attach a carbon fiber extension to the existing pole, if a lighter camera in a higher position is desired.

The decision was finally made to move ahead with vlogging gear. While a wireless headset was being planned for years, a more practical system was a wired headset attached to a phone worn by the athlete. The phone would power the headset over USB & receive audio over the headphone jack.

There is no commercially available headset which captures decent audio, is wireless, & can monitor itself. The best commercial solution would use FM radio to transmit &...Continue Reading
Posted by Jack Crossfire | Nov 18, 2017 @ 11:37 PM | 3,335 Views
Anti stiction was implemented for this one. The trick was determining when the steering needed to change direction & kick the servo. The zero crossings of the P error were a good indicator of when the steering needed to change direction, although I & D also factored into the servo direction. The decision was made to rely on P error so the algorithm wouldn't feed back into itself. It waited until the zero crossing + a little hysteresis, then added a fixed amount to the I term to kick the servo. Using 1 degree of hysteresis & 3% for the kick completely solved the stiction problem. Not using any hysteresis or using too much kick made the steering oscillate. It was so tight, it could stay on the path for a nearly quarter mile with no steering inputs.

Also, the transmission was lubricated through the holes for the 1st time. The tape didn't stop the oil from leaking out, but hopefully it kept dirt out. Oil leaked out the bottom. It was so quiet after lubrication, you could hear the servo.

The speed was set to 10min/mile. Burned 3886mAh +4477mAh or 331mAh/mile. This was finally near the predicted range, but manely from increasing the speed. The lion can't manetain that speed for every run.
Posted by Jack Crossfire | Nov 18, 2017 @ 03:05 PM | 3,584 Views
In modern times, there's no need for bodging a webcam on an optical microscope. Documentation is better achieved with a DSLR macro lens. The next step up is a cheap USB microscope. The working distance is too short on the Amscope. The angled light is dim & creates too many shadows. The incandescent lightbulb gets flaming hot. It takes too much space. Its only redeeming factors are it's stereo, it has a higher framerate, & it was free. Soldering with the lag of a USB microscope & without depth perception might be challenging.

This microscope came in handy for soldering, but never for documenting. The webcam experiment was a rite of passage. You can get decent photos by holding a phone cam exactly the right distance from the eyepiece. It won't work with a DSLR.

Webcam on a 40x microscope (2 min 49 sec)

Posted by Jack Crossfire | Nov 16, 2017 @ 01:17 AM | 3,493 Views
After another 25 miles of listening to transmission noise, the decision was made to tear it down again & relube it. The counter gear was the only dry gear. Drilled holes to make it easier to relube without tearing it down. The mane problem is plugging the holes.

The 2nd 25 miler burned 4036mAh + 4605mAh or 345mAh/mile. It was much lower because the speed was increased to 10:30/mile. The motor got more efficient when it got faster. The lion also got more efficient, burning only 1 Clif Blok instead of 2.

The next problem is steering deadband. A lot of play from the servo saver makes it drive erratically. That was a buster to jockey for 5 hours. The servo needs to kick it when it reverses direction or it needs knowledge of the rod position.

There's increasing P gain & adding extra offset to I when the steering reverses direction. Steering constantly reverses direction by 1. Finding the minimum amount of reversal to be considered a real reversal would be tricky.
Posted by Jack Crossfire | Nov 12, 2017 @ 01:39 AM | 3,333 Views
Finally achieved 25 mile range on a vehicle. It only carried food for the lion & the headlights ran for 2/3 of it. The right front tire interfered with the new battery clamp, but it was otherwise flawless. Brushless motor, wheel alignment, ball bearings were key to getting this range on the lunchbox.

It burned 4370mAh +4907mAh or 362mAh/mile, so well above the predicted burn rate. The transmission oil may have worn out. To feed the lion, only 2 Cliff Bloks were required. The speed was reduced to 12min/mile, to aid the lion. That reduced the animal burn rate while it may have increased the machine burn rate.
Posted by Jack Crossfire | Nov 10, 2017 @ 11:45 PM | 3,343 Views
The battery popped out for the 1st time in 3 years while running down some hills at 10mph.
Posted by Jack Crossfire | Nov 07, 2017 @ 12:08 AM | 4,014 Views
It was always the same driver kit available anywhere in Shanghai for a buck, & the only way this many bits could be under $25 is if they were made of the softest steel & would only last for 1 teardown, if you were lucky, but the ifixit packaging was 1 of a kind. The magnetic case with grid to organize your screws, perfectly angled bits, not perpendicular but angled, perfectly sized foam insert just exudes the kind of quality that couldn't be a bog standard Shanghai street hustle, but something hand crafted just for the kind of user whose laptop looks like a work of art with an Apple logo.

To be sure, the only reason anyone bought this kit was the #5 pentalobe bit to fix his Macbook Pro. While they've always marketed their kits as aftermarket, there really is no other supplier, so it wouldn't be surprising if the genius bar used the same kit....Continue Reading
Posted by Jack Crossfire | Nov 04, 2017 @ 06:43 PM | 3,281 Views
The 1 week order delay finally ended & the journey to replace all 9 bearings began, manely rebuilding the transmission. The counter gear & the rear wheels were finally replaced.

A 13.2 miler started out super quiet & got slightly louder after 12 miles, as it was with the previous transmission rebuild. It could be more easily relubricated with WD40 to maximize the range. The evidence is the bushing inside the counter gear was causing most of the noise, since it got super hot & spun super fast.

With no payload but a speaker, headlights on, & hilly terrain, it used 2065 + 2305mAh or 331mAh/mile. A new record low consumption which could go 27 miles, but with no payload. With cold weather & no need to carry a shirt, the 27 mile range could be a realistic goal. A lion can only manetain that speed for 16 miles, so at a reduced speed with a payload, we could have a vehicle reach the end of the valley for the 1st time. The counter gear bearing must have been sucking a lot of power.

Did another 6.75 run with it. The 10mph test revealed no material improvement in steering oscillation. Noise was still reduced, but not as much as the Sandstorm. Despite headlights, 13 miles of wear on the lubrication, 3km at 10mph & a return trip with 10oz of food, it still only burned 2215mAh or 318mAh/mile.