|
OK, fine, I will do my own A/C. I think I know what's wrong, too. There's a scratch/burr on the inside of the tube, near/at the o-ring seal. Professionals looked at it and thought it wouldn't be a problem, but I have doubts. And speaking of leaking seals, my rear shock is coated in oil, gently caress! I replaced it only 18 months ago and it can't have more than 25k miles on it.
|
#
?
Oct 8, 2022 09:57
|
|
|
|
| # ? May 11, 2024 10:17 |
|
|
ryanrs posted:OK, fine, I will do my own A/C. I think I know what's wrong, too. There's a scratch/burr on the inside of the tube, near/at the o-ring seal. Professionals looked at it and thought it wouldn't be a problem, but I have doubts. 25k of your miles are like 100k "normal" miles.
|
|
#
?
Oct 8, 2022 13:23
|
|
|
sharkytm posted:25k of your miles are like 100k "normal" miles. yeah, off-road miles are not normal miles! unless this was a deliberately highly specced off-road shock you totally got your moneys worth out of it!
|
|
#
?
Oct 8, 2022 14:02
|
|
|
A year and change out of a normal shock dealing with conditions it wasn't really designed for is quite reasonable imo.
|
|
#
?
Oct 8, 2022 16:58
|
|
|
=== Project(s) Status === Previous repair list from 7 months ago. It was slow going since I had to learn to weld and setup a metalworking shop to do some of the repairs. Only two tasks remain, the skid plates and the a/c delete, and those will be done in a week. Hopefully this next batch of modifications will be completed faster (except the anti-bear sensor net, which is a more open-ended). Sorted in approximate chronological order. Skid Plates v2 I'm currently painting the skid plates. They are getting a heavy coat of Rustoleum Hammered, a thick, goopy oil paint. I'm going to let the paint harden for a week or so before putting the finished plates on the van. Rear A/C Delete More than a success, not only have I deleted the rear A/C, but also the front. It sounds like I might tackle this myself instead of going back to the shop? The fitting that's leaking has an o-ring seal, and the cap has a scratch in that area. I think I can polish it out pretty easily. Offroad Light Switches This is a bigger project. I know how to do it, and what parts I'll use. I just need to design the circuit board and mechanical stuff. Which I guess is everything? I just need to design everything. I am including a certain amount of automation. For example, the forward-facing lights, when on, follow the stock highbeams. So dimming the headlights turns off the offroad lights. Similarly, I'm going to put in a bright reverse light that turns on when the cornering lights are on and I shift into reverse, etc. Originally the switch panel was going to be 100% hardware, no microcontroller. But I have a funny idea for an anti-theft blinking LED, and it's much easier to implement the pseudo-random blink pattern in software. So I added a little microcontroller. Of course, now that the microcontroller exists, there is the urge to let it control everything, not just the blinking LED. Backup Camera The van has totally poo poo rear visibility, especially on uneven terrain. Very often the only thing you can see out the back is sky. I have the camera and screen, I just need to install them. It'll be easiest to do this at the same time as I install the light switches. Bear Alarm I want a bear early warning system to alert me if a bear enters my camp at night. Then I can chase it away before it gets into my food and destroys everything. I'm guessing bears spend a few minutes quietly poking around before digging in, so that's my window. I made a prototype networked radar / thermal smart sensor, but I still need to write a bit of firmware to verify it actually works. I want to go full sci-fi / heist movie with this tech. For example, the system should be smart enough to draw a map of the sensor deployment automatically, without me having to gently caress around in ms paint each evening. Biggest risk is the software, because I will probably lose interest when it's only 10% complete. Really not looking forward to making an iphone app. M18 Ham Radio I want to make a power converter so I can run my 50W ham radio off my M18 chainsaw battery. I want to be able to take the radio out of the van, and hike it to the top of a ridge to get reception. Rear Shocks At least one is leaking oil. I need to look into it sometime, but it's not urgent (yet). I am currently using KYB shocks, which are popular for Siennas. Is there a shock that would be more durable that still fits the van? What do I look for? e: KYB Gas-a-Just Shock what I'm currently using. They are quite cheap at $37, and pretty easy to replace. Maybe just keep on replacing them every 18 months forever? Gas Tank Skid Plate I should probably make one to protect my new gas tank. ryanrs fucked around with this message at 05:52 on Oct 9, 2022 |
|
#
?
Oct 9, 2022 05:40
|
|
|
Trying to parody the TRD design, but needs some work. KiCad pcb layout tools may not be up to the task, so I might have to learn Inkscape.
|
|
#
?
Oct 10, 2022 07:37
|
|
|
Light Switch Design For the mechanical design, I'm thinking of two parallel circuit boards, spaced 0.5" apart. Toggle switches will be mounted cordwood style. The switch terminals will be soldered to the rear circuit board, which will also hold all the circuitry. The front circuit board won't have any circuit traces on it, just round mounting holes for the switches. It's structural, not electronic. I could make that front panel out of laser-cut stainless (for example), but if I use a pcb, then I can take advantage of the high-res photolithography process for printing the switch labels, art, etc. I'm thinking of using the OSHPark After Dark board stackup. It's a basic 2-layer board, with an opaque black core and clear solder mask. Pics of After Dark pcbs. The palette is black, copper, gold, white. I'd like to incorporate a copper-to-gold gradient somewhere in the design, made with half-tone dots on the solder mask. Showcase some real abuse of the pcb cad tools, heh.
|
|
#
?
Oct 10, 2022 08:24
|
|
|
Bear Radar   Some pics of the first hand-soldered prototype. The small blue board holds the CPU, radios, and i/o. It plugs into the back of the green board, which has the thermal motion detector, radar motion detector, other sensors, and and LED floodlight. Enclosure is 3D printed with HP multijet fusion. Specs: Nordic nRF52840 ARM CPU Bluetooth LE 900 MHz 1W LoRa radio BNO055 orientation sensor (up, north) U-blox MAX-M10 GPS 24 GHz doppler radar Passive IR motion detector The blue board has most of the expensive chips, and not too many passives, so I will probably hand-solder all of those. It will also take some time to get the radio stuff tested. The green board, on the other hand, is a pain in the dick to hand-solder because it has a pile of 0402 passives. The leadless LEDs suck, too. So once I finish proving this first prototype works, I will hire a Chinese contract manufacturer to make 10 copies. I hope to get that process going sometime this month. In the future I may make more devices, using the same blue CPU/radio board, but mated to a peripheral board with different sensor packages, other output devices, etc. One idea is a speaker to play various barking dog sound loops, controllable through the radio network. With a handful of nodes and some basic spatial audio poo poo, you could simulate a pack of dogs circling the campsite, ha ha. While not strictly car-related, this project is interwoven with my minivan trips, so you'll hear a bit about it. The software side will stretch out 1 year+. In the beginning it'll just be a networked motion detector. But I hope to add cool stuff like self-mapping sensor deployments at some point.
|
|
#
?
Oct 10, 2022 23:10
|
|
|
kastein posted:E: check out Motronic's AI AC thread for a basic introduction. You technically need an EPA license to do this but if you're only working on friends and families cars and your shitboxes it's very likely it'll never come into question. The EPA Section 609 (Mobile AC) license is easy to get - I've had one for something like 25 years, took the test online - and cheap (currently $24.) I *think* these were the folks I used: https://macsmobileairclimate.org/about-609-certification/ Scroll down the page for training materials, etc. They have a $35 for the video webinar, but the textbook is plenty - that's all they had when I did it. Test is open-book (and easy if you actually read the book and have half a brain.) PDF textbook: https://macsprod.freetls.fastly.net/wp-content/uploads/2021/04/2021-609-Certi-Train-Man-English-LR.pdf
|
|
#
?
Oct 11, 2022 00:20
|
|
|
I'll have to do that, thanks. Ryan, holy crap, that's a fairly involved project. I like it. BTW I've ordered After Dark boards before for a few projects and they look great. Their pricing is pretty solid for what you get.
|
|
#
?
Oct 11, 2022 02:10
|
|
|
ryanrs posted:Bear Radar Goddamn this boggles my mind. When you posted the idea for this upthread I thought it was a joke you slipped in there! Some real Snow Crash poo poo, you should get it branded Ng Security Industries 
|
|
#
?
Oct 11, 2022 02:37
|
|
|
Mouser made me file an End-User Certificate for the thermal motion detectors, ha ha. It's the exact same sensor as used in motion-detecting outdoor floodlights, but I had to promise I wasn't going to ship it to Putin. Which is fine. gently caress Putin, no automatic landscape lighting for him!
|
|
#
?
Oct 11, 2022 03:31
|
|
|
You're an absolute madman. I love it.
|
|
#
?
Oct 11, 2022 04:32
|
|
|
ryanrs posted:
Would it be easier to deploy the sensors on the corners of your van? That way, you get proximity alarms around where you're sleeping and you don't have to re-jigger the map of how they're deployed. Also, obviates the need for an iPhone app if it just flashes some lights and sounds inside the van.
|
|
#
?
Oct 11, 2022 20:09
|
|
|
Safety Dance posted:Would it be easier to deploy the sensors on the corners of your van? The device needs to be reasonably self-contained and compact, so I can show it off to interested parties such as potential employers. That's also why I am taking pains to design-in interesting hardware capabilities, but I'm not fully committed to actually implementing the software. If this was an actual consumer electronics product, the app-level software and network code would be done by a different team. It's too big for one person to do it all in a reasonable amount of time. I specifically want to show I can take a hardware design from concept, through circuit design, mechanical design, firmware, and outsourced manufacturing (just board assembly and enclosure 3d printing for now). Scaling up manufacturing beyond 10-100 prototypes is also a job for other people (probably need to re-design/optimize everything at that point, e.g. injection molding). So that's the background context for this project.
|
|
#
?
Oct 11, 2022 20:56
|
|
|
ryanrs posted:The device needs to be reasonably self-contained and compact, so I can show it off to interested parties such as potential employers. OK, yeah, fair enough. It's an interesting tech demo!
|
|
#
?
Oct 11, 2022 21:07
|
|
|
Does it just detect bears or other critters? Like will raccoons set it off? Deer? People? I can see raccoons being annoying if they set it off but if you strap your cooler closed with say, a ratchet strap or two they ain't getting in there. A bear is different of course cause they don't give a gently caress about straps
|
|
#
?
Oct 12, 2022 00:29
|
|
|
TBD. But I have recipes for raccoon, and a girl I know has asked me for a pelt.
|
|
#
?
Oct 12, 2022 02:59
|
|
|
Gas Tank Skid Plate This is probably my lowest priority project, and it's still in the brainstorming phase. But I'm pretty sure I don't want to make it out of mild steel. The design time and fabrication time is so much greater than the cost of materials, that it seems silly to use A36 steel, which is the cheapest steel there is. I should be making my skid plates out of A514, Grade 80, or any of dozens of proprietary high yield strength steels. I could use 4130 chromoly, but I don't want to do post-fabrication heat treatment, so it would be in the normalized state, so about 63ksi yield strength, double that of mild steel. Which is not as strong as more modern AHSS, plus chromoly has way more carbon, and I think it's more expensive. Really what I want is some of the new steels being used in vehicles. Something like McMaster-Carr UHSS 0.159" would be great, though I need larger sheets. I think that McMaster steel is Tata Ympress 100 XF. Along with stronger steel, and I also want to bend it, to create stronger shapes. To that end, I spent the afternoon poking around local metal fabrication shops, looking to see who had the necessary equipment. Keller Industries has a huge waterjet cutter. They will be able to cut anything I need. AG Heagney next door has 65T and 130T press brakes, which are probably about the right size for what I need. They are giant monster machines. E. M. Jorgensen in Hayward has a lot of steel alloys. I need to ask them about 100 XF, as it has less carbon than typical A514, and is more formable. I'm still looking for other steel vendors, though. That's my vision for advanced skid plates: more strength and less weight by using more advanced steels. Modern alloys are so, so much better than plain old mild steel. I don't think it's even that much more expensive than mild steel (although it's hard to find published prices). It can't be crazy expensive, if they're using it to make truck frames.
|
|
#
?
Oct 12, 2022 05:18
|
|
|
I know a lot of things like excavator buckets are made out of AR500, maybe that? The guys in my press brake fb group seem to find it fairly difficult to bend.
|
|
#
?
Oct 12, 2022 05:23
|
|
|
Yeah, AR500 is not as formable, needs larger radius bends, and is harder to weld. It feels like it is too far in the other direction, too hard and difficult to work with.
|
|
#
?
Oct 12, 2022 05:52
|
|
|
Check out KVA's MS3. They're in SoCal.
|
|
#
?
Oct 12, 2022 06:00
|
|
|
I need ~0.16" plate.
|
|
#
?
Oct 12, 2022 06:50
|
|
|
ryanrs posted:The green board, on the other hand, is a pain in the dick to hand-solder because it has a pile of 0402 passives. The leadless LEDs suck, too. So once I finish proving this first prototype works, I will hire a Chinese contract manufacturer to make 10 copies. I hope to get that process going sometime this month. IIRC, you're in the Bay Area? I'm in Mountain View and have a reflow oven (Whizoo Controleo3, pretty reliable) if you need to borrow one. IMO the nRF52840 and that stupid loving two row QFN package is way worse than 0402s...
|
|
#
?
Oct 12, 2022 08:33
|
|
|
kastein posted:I know a lot of things like excavator buckets are made out of AR500, maybe that? The guys in my press brake fb group seem to find it fairly difficult to bend. You would be in a press brake facebook group.
|
|
#
?
Oct 12, 2022 17:34
|
|
|
The Linux Fairy posted:IIRC, you're in the Bay Area? I'm in Mountain View and have a reflow oven (Whizoo Controleo3, pretty reliable) if you need to borrow one. IMO the nRF52840 and that stupid loving two row QFN package is way worse than 0402s... I'm using a tiny dev board right now, so that weirdo QFN is 2 steps removed from anything I have to solder: 1. My blue board. 2. Adafruit ItsyBitsy nRF52840 Express super easy to solder. 3. Raytac MDBT50Q-1MV2 (which has its own cursed footprint, but also FCC approval) 4. Nordic nRF52840 MCU. A future board shrink will probably use the Raytac module directly. For my design, I think that is the sweet spot. ryanrs fucked around with this message at 18:08 on Oct 12, 2022 |
|
#
?
Oct 12, 2022 17:45
|
|
|
ryanrs posted:I'm using a tiny dev board right now, so that weirdo QFN is 2 steps removed from anything I have to solder: Makes sense. Well, if you need anything reflowed, let me know. I basically never use it anymore, other than to bake SLA resin before curing it...
|
|
#
?
Oct 13, 2022 06:41
|
|
|
Thermal Vision The IR motion detector might have an issue. It takes many minutes to stabilize after power-on, during which it is unresponsive. This is fine for out in the woods, but not great for a demo. I'll add an alternate set of solder pads under the IR motion detector, so I can swap it for a 2D focal plane array if needed (aka a real thermal camera). The Melexis 90640 is a tiny thermal camera module. It is moderately expensive ($75) and very low res (32x24 pixels = 768 pixels total). Because it's so low-res, every image looks like a picture of a ghost. It's perfect! 
|
|
#
?
Oct 13, 2022 19:03
|
|
|
Is there perhaps a low power standby or sleep mode you can put it in on the way to the demo so that it's not killing the battery but is ready to rock the second you take it out of standby? A micropower or high efficiency regulator for it plus that might do the trick. Or possibly lower its clock to minimize power use while it's stabilizing? I haven't read the datasheet, but these are methods I've had to use before.
|
|
#
?
Oct 13, 2022 21:22
|
|
|
No, it's even weirder, and likely just a bug(s). Even after sitting for a while, the sensor behavior was somewhat erratic. I started following traces and realized I had a broken net / floating VDD. Classic CMOS running through the ESD diodes, easy fix. The second fix, which was also necessary to make it work, was enabling a 100k pullup to the PIR output signal. But nothing in the sensor docs suggests the output is open collector (OC). None of the given example circuits should work with OC. Even the TM2291 chip datasheet doesn't say anything about OC. So either the Chinese docs are total dogshit and nobody ever even tried the example circuits (lol completely possible), or perhaps I damaged the output drive while swapping PIR sensors (I tried two, but did the swap before fixing the floating VDD, so not 100% conclusive). Not an entirely satisfying conclusion, and maybe needs more scrutiny. But with these two changes, the IR motion sensor is now super sensitive, no false triggers, and is completely well-behaved. Hmm.
|
|
#
?
Oct 14, 2022 01:55
|
|
|
Heh, the doppler radar reads 'target approaching' when placed in front of a fan, and 'target receding' when pointed at the back. Faster fan speeds give higher km/h speed readings. This is a scenario when the thermal + radar sensor fusion can prevent a false alarm. If the radar sees some low speed motion, too slow to be a vehicle, and the thermal sees nothing, then maybe it's just the wind blowing branches (or a ghost).
|
|
#
?
Oct 14, 2022 02:46
|
|
|
 I got hosed by CMOS floating pins once years ago, turns out when you don't tie #WE and (IIRC, it was like a decade ago) Vpp high on a 28Fxxx series EEPROM, it will test fine at low speed but over a few MHz you get the same data for like 4 bus cycles in a row no matter what you asked for. Whoops.At least it was an easy fix once you found it.
|
|
#
?
Oct 14, 2022 05:33
|
|
|
Power Strategy In order to reduce the scope of this project, I took some shortcuts that would not be possible on a real product (probably). A big compromise is that my Bear Radar does not have an internal battery, and therefore lacks all the battery boilerplate circuitry: battery charger, battery protection, battery gauge, etc. I decided to skip all the battery complexity and power the device with a USB power bank. But before I committed to that path, I bought a few power banks off Amazon to explore their performance, in particular quiescent current. Spoiler: this $22 Anker power bank has significantly better performance and build quality than the no-name ones. Most/all USB power banks automatically shut off if you don't draw significant current (100+ mA). This is because the power bank only wants to charge your phone, it does not want to supply idle power after your phone is done charging. There are good UX reasons for this behavior, though it is slightly inconvenient. To keep the Anker power bank alive, you need to draw 150 mA for 1 sec, every 100 seconds or so. Easy enough if you have a mosfet and power resistor on one of your GPIOs, which the Bear Radar does have. But because of this auto-shutoff behavior, there is not much incentive for designers to optimize their power banks for quiescent current. So I devised a "power bank minimum load" for testing. I used a Teensy 4.0 with no external hardware for this test. Every 100 secs, the CPU wakes up and runs a busy loop for 1 sec. I had to overclock it to 816 MHz to draw enough current to wake the power bank, ha ha. Meanwhile, the little red LED on the Teensy is kept lit 100% of the time. So that's the test: How long can a power bank keep a tiny LED lit, while using the minimum current to keep it awake, power the 3.2 to 5V step up converter, etc. The answer for that Anker power bank was over 2 weeks before running out of charge. Thanks, Anker engineer, for caring about quiescent current! Power banks with legit good electronics inside: Anker 10k slim and Anker 10k Slim USB-C. How Thirsty is the Bear Radar? Measured current consumption, no optimization yet, though the Arduino lib does a low power sleep while waiting on delay(ms>=2). 3 mA Passive IR only 10 mA IR + Radar in low duty cycle mode (scans at 1 Hz) 70 mA IR + Full power radar (or low power radar when it sees something) Other power draws not yet measured: - GPS, but we don't need to run it continuously, not in motion. - BNO055 IMU/orientation sensor, guessing it has good lower power modes. - Radio, yeah this one will use 20 mA or so if it's always listening. I have some interesting weird/bad ideas for the radio that may preclude powering down the receive section. So I am budgeting for 100% RX current, rather than low duty cycle channel monitoring. The core radio is a Semtec SX1272 with a 1W PA of dubious legality (not that anyone will notice on 900 MHz). LoRa modulation, but not LoRaWAN protocol. More like Meshtastic. The LED floodlight can draw as much current as you allow, so I've capped it just under 2.4A. You'll have to dim the LEDs a bit if you want to transmit a packet at full power, though. Or maybe not? I wonder how much current the Anker can put out for 5 secs without getting mad? My target battery life is several days, a week would be great. The power figures above make me think I can definitely hit this goal. Actual battery life will depend on how chatty the radio protocol is, required latency, etc. I'm not chasing max battery life, I just don't want it to be terrible.
|
|
#
?
Oct 14, 2022 08:39
|
|
|
Now I'm wondering about how much heat a bear gives off at different ambient temperatures. Interesting project, you can always optimize power consumption in version 2.
|
|
#
?
Oct 14, 2022 14:40
|
|
|
BEDAR® (BEar Detection And Ranging)
|
|
#
?
Oct 14, 2022 14:44
|
|
|
SpeedFreek posted:Now I'm wondering about how much heat a bear gives off at different ambient temperatures. Interesting project, you can always optimize power consumption in version 2. Take it to the next level, take a FLIR down to the zoo on various weather days. EDIT: Also dress up as the Predator.
|
|
#
?
Oct 14, 2022 15:09
|
|
|
Halloween is coming but the local zoo uses glass and not bars for the animals that can eat the guests. This looks like a better design than the driveway alarm that can't differentiate between a chipmunk and UPS truck.
|
|
#
?
Oct 14, 2022 16:17
|
|
|
I tried searching for the radar cross section of a black bear, but I can only find stuff about the Tu-95. e: I just realized that this glorified motion detector floodlight ought to have an ambient light sensor on board. Adding a VEML7700. ryanrs fucked around with this message at 22:09 on Oct 14, 2022 |
|
#
?
Oct 14, 2022 18:40
|
|
|
This may impact bear radar chip production. https://twitter.com/arstechnica/status/1580734647147048961?s=20&t=RjffiSB2DwMo7zodGJbNUg
|
|
#
?
Oct 15, 2022 03:43
|
|
|
|
| # ? May 11, 2024 10:17 |
|
|
Lol if you don't buy enough chips for a full manufacturing run before you even design the pcb. I've been gathering chips for this project since July. There are surprisingly few Chinese chips in the Bear Radar. I avoid them because the documentation usually sucks. e: One example is the GPS. I'm using EU u-blox GPS modules. But the u-blox MAX-M10S was out of stock a few months ago, so I took a look at some pin-compatible Chinese GPS modules made by Quectel. But even those cheaper Chinese modules aren't based on Chinese chips, the GPS processor is from Mediatek (Taiwan). The Chinese LoRa radio is based on a Semtec RF chip (California). The microcontroller dev board (US) contains a MCU module from Raytac (Taiwan), which has a Nordic (Norway) nRF52 microcontroller based on ARM architecture (Japan...?) and is fabbed on TSMC's 55ULP process (Taiwan). When I send the Bear Radar BOM to the Chinese factory, all the important components will be provided by me as a 'kit', for the Bear Radar this includes all chips, discrete transistors, connectors, LEDs, and inductors. The boring components, like the resistors and capacitors, will be specified by value, tolerance, temperature rating, size, etc, and the factory will supply them. ryanrs fucked around with this message at 06:50 on Oct 15, 2022 |
|
#
?
Oct 15, 2022 05:02
|
|