Jump to content

esaj

Moderators
  • Content Count

    4,546
  • Joined

  • Last visited

  • Days Won

    135

Everything posted by esaj

  1. I calculated the values for the LTC3638 simply following the datasheet and using a spreadsheet, unfortunately I haven't figured so far how to get it to Google Sheets without breaking the formulas or values changing (for example, we use comma as decimal separator here...) LTC3638_calcs.ods The cells with cyan backgrounds are for input values, rest is calculated based on those: Reason for picking the non-synchronous LTC3638 vs. synchronous LTC3639 was that the synchronous-chip is limited to 100mA max (if it's an absolute maximum, shouldn't draw that much, otherwise risking breaking the device). It might be that the display + SoC + other stuff doesn't need but 50mA, but in case of using a different display or such requiring more current, the 3638 has 250mA maximum, leaving much more headroom. I was thinking of using 5V output, but it could be set somewhat lower, the final 3.3V voltage should be regulated through a linear regulator anyway to filter the noise in the measurement circuits, or if 3.3V is taken directly from the SMPS, the analog-side may need separate filtering. Running an LTSpice-simulation with component values taken from those below (the preliminary parts) and 101V input, the ripple at constant steady 100mA or 50mA draw was only around 25mV, but it's just a simulation, not taking into account the effects of layout for example (current loops around the SMPS), and the actual load has an MCU, which draws current in "spikes" with the clock, although the bypass caps of the MCU should filter out most of it. Also the simulation shows that the output voltage raises above 5V with lower load, so using 3.3V might be a no-go if it does the same (the MCU can't handle but 3.9V absolute max), so probably higher voltage than 3.3V at the SMPS + post-regulation is needed anyway. Isolated flyback might be a better option than "normal" buck-regulation, but I don't have experience in using such, and I doubt this project would be a good place to start learning Seba likely knows much better, so probably best to wait until he's got at least a preliminary design. I won't order the parts for the step-down yet, as most of the the circuitry can be bench-tested with a separate power supplies, and for load testing I'll need something more powerful anyway, my adjustable 0-100V max linear PSU is only 100W anyway (limited to 1A max at any voltage, so it can only reach 100W with 100V output). At least it does have ability to push the voltage up to 120V for a short while (+-20% buttons), for line transient / overvoltage testing. I can start poking into the actual measurement circuits using what I have at hand (General purpose Cortex-M's without BT, op-amps, current sense resistors, displays...) in the mean time. Preliminary parts for the step-down: (amount) (manufacturer) (part number) (specs) (Mouser prices in € without VAT: x.xxx < 10 pieces / x.xxx 10-99 etc., the price breaks) Diode: 1-2 ST Microelectronics STPS1170AF 170V / 1A / 0.62Vf / SMAflat 0.384 < 10 / 0.299 10-99 / 0.12 100-999 2 if using the same diode for input protection before the step-down Inductor: 1 Laird Performance Materials TYS6045680M-10 68uH 20% Shielded / 1A / 289mOhm / 6x6x2.8mm 0.297 <10 / 0.286 10-24 / 0.268 25-49 / 0.261 50-99 Input caps: 3 Yageo CC1206KKX7RABB104 100nF / 200V X7R MLCC / 20mOhm @ 40kHz / 1206 0.262 < 10 / 0.177 10-99 Might also be 220nF, haven't checked options further yet, likely the input capacitance should be made higher than this to minimize input ripple (the Delta_Vin used in the spreadsheet is pretty large at 5V... ), more like a couple of uF? SS-cap (100nF actually makes the soft-start last 16ms): 1 Murata GCM155L8EH104KE07D 100nF / 50V X8L MLCC / 300mOhm @ 40kHz / 95nF @ 5V Bias / 0402 0.087 < 10 / 0.071 10-99 100nF's will likely be needed all around (bypasses to MCU, amplifier, linear reg...), so the actual amount is more, something like 10. Output caps: 2 Murata GRM21BC71C106KE11L 10uF / 16V X7S MLCC / 10mOhm @ 40kHz / 6.3uF @ 5V Bias / 0805 0.384 <10, 0.269 10-99 / 0.176 100-249 1 Nichicon RNE1C101MDS1 100uF / 16V Polymer / 35mOhm 0.533 <10 / 0.396 10-99 / 0.272 100-499 LTC3638 7.72 < 25 Resistors: Basic 1% is probably accurate enough for UVLO/OVLO (have to take into account the voltages with package sizes, ie. clearance / creepage, 0805's seem to be usually rated at 150V max, so maybe 1206's)
  2. Why not, but it doesn't make it much more complex to add the current monitoring and thus the ability to calculate watthours going into the battery. In both cases, the MCU will still need around 3.3V voltage, which needs to be stepped down from up to a 100V and change (even slight overvoltage).
  3. If you need a reliable low voltage lab PSU in the future (for guitar pedals or whatever), look no further than HP6632 (A or B), which are relatively easy to find as second-hand, this is a serious industrial-grade 100W programmable linear lab rack PSU from the 80's/90's (production discontinued in 2017) with proven track record. East Trade Promotion sells such for example (not sure if they have any in stock right now, but there have been plenty in the past), I bought mine from them through huuto.net: http://sivut.klikkaa.fi/eastrp.fi/webshop/product_details.php?p=45 https://www.keysight.com/en/pd-838596-pn-6632B/100-watt-system-power-supply-20v-5a?&cc=FI&lc=fin (The reason the specs are in Keysight-website is that HP instrument-division first became part of Agilent and then Keysight) Output Ratings Output voltage: 0 to 20 V Output current: 0 to 5 A Programming Accuracy at 25°C ± 5°C Voltage: 10 mV + Current: 0.05% + 2 mA Ripple & Noise (20 Hz to 20 MHz) Voltage Normal Mode rms: 0.3 mV Voltage Normal Mode peak-to-peak: 3 mV Fast mode rms/ peak-to-peak: 1 mV/ 10 mV Current rms: 2 m https://www.eevblog.com/forum/reviews/hp-6632a-20v-5amps-programmable-psu/ "They are super precise , and can sink & source current.The meters are/were the most accurate i had when i bought them." Mine was off by about 4mV (0.004V) higher up in the voltage, that's about it. Hasn't been calibrated in 10+ years. In a pinch, it'll work also as a load up to 20V/5A, although at least on mine the down-programmer has an issue where the sinked current will be about 250mA higher than what's programmed, guess it's an general issue (it's not like it was meant to be used as such ): https://www.eevblog.com/forum/testgear/hp-6632-current-sinking-performance/ Not that big of an investment money-wise (I think I paid 120€ for mine with rack-ears, including shipping), the size might be an issue (2U rack-unit), plus it weighs 10+kg.
  4. Oldish topic, but I decided to play a necromancer.. How have you guys found a big change in tire pressure changes the ride behavior? I've favored high pressure (over 4+ bar / 58+ PSI) due to how much "faster reacting" it makes the wheel, not to mention that I get much better mileage at high pressure (can't quote exact numbers, but I'd say 1/3 to 1/4 more mileage vs. really low pressure, like 2-2.5 bar / about 30-36 PSI). Now, I've tried different pressures a long, long time ago with the Firewheel, and found out that I favor the high pressure over lower pressure (especially since my FW had such a small battery at 260Wh, more mileage ), despite making balancing at any speed much more delicate. Usually I've topped up the pressure after two-three months, but this time it seems my tire pressure had been dropping faster than usual. I think I've also been riding faster than usual lately, with average speeds topping 30km/h (yes, yes, I know it's not "that fast", but it's a KS16S ). I'd consider myself a fairly experienced rider (but there are more experienced in the forums) after something like maybe 10k+ km behind me (I don't know, I stopped measuring a long time ago, but then again, I only get to ride 5-6 months per year, since spring of 2015). So, early this week I noticed my tire pressure is getting really low, since I could just "bounce" on the wheel (can't say exact figure, since my car-battery powered -compressors' meter doesn't seem to even react to anything below 2.5bar or thereabouts), and was worried I could cause a "pinch puncture" riding over some corner stones or whatnot. The bright side was that slight irregularities on the road felt like nothing, and I could ride fast into turns with no issue (although at some point I likely would have hit the pedal on the ground). I pumped up the tire until the compressor said something like 4.3-4.5 bar or thereabouts. I pump the pressure "over", because some of the air always escapes removing the compressor nozzle and the valve extension, and the final pressure could be more like 3.5-4.0 or something like that. Probably should invest in a pressure meter to know better Also the cheapest battery-powered compressor in the hardware store probably doesn't have the most exact measuring. Anyway, this time I managed to rip out the nozzle and extension really fast, with very little loss in air. Of course I didn't test the wheel afterwards, as it was late in the evening, just went to sleep. The next morning when I rode to work, holy hell, was the wheel acting strange. The tire was super hard, I've used to feeling small pebbles and such on the road, but now it was like the tire had zero give at whatever slightest irregularity the road might have. The wheel was super sensitive to even my slightest move, and would start to wobble on its own riding straight on a more or less even road. But the biggest surprise was the effect on cornering. I'm used to riding pretty fast into 45 or even 90 degree wider turns, when there's not people around and I can see that I'm not going to hit anyone after the turn. Also, I'm used to leaning more on my body, and only more slightly tilting the wheel on turns, like "leaning out" from the wheel, allowing the foot opposite to the turn to lift off a bit from the pedal, remotely what you'd do on a motorcycle pushing your knee out and "hanging" from the bars, leaning out from the bike, but this was much different to before. The wheel seemed to stay more upright, like totally upright, and resist me trying to tilt it with my calf. This was at speeds something like 25-30+km/h (first warning playing) and change, tried it a couple of times. After a 5-6 kilometers, I had to stop and let some of the air out to ride rest of the way. I don't know where I dropped it, likely something like 3.5-4 bar, since it's more like usually after filling up the tire. What most surprised me was the cornering behavior, since the wheel "fought back" on leaning sideways a lot more than what I remember having experienced before. I'm a lightweight rider (probably pretty close to 60kg/ 130lbs in full gear), so heavier riders might not have the same effect, but it certainly made my turns much wider than usual at higher speeds. In lower speeds, I could still ride fine at sub-walking speeds, while simultaneously opening my helmet and lighting a cigarette, so at low speeds it seems not to be much of an issue. Might still be that I'll start to prefer using lower pressures (but not too low ), considering that allowing the pressure to drop, it made fast speed riding far more stable ("you haven't lived until you make a tight 90-degree turn with full tiltback"... nah, actually, I don't recommend it ), and riding at slug pace was even easier than with high pressure. Of course on the low end, you're going to likely hit a point where it becomes an issue (I was already noticing that I was using more battery than usual and probably would have hit the pedal on the ground sooner or later on a tight turn).
  5. Was going to write about the same thing, but you got it first... Yeah, the 24S packs will have higher internal resistance, couple that with less packs in parallel, it might be that the 100V versions actually have lower maximum (safe operating range) power output vs. 20S, and the only upside is (up to a point) higher top speed. But there are a lot of variables at play, so it might well be that the the internal resistance of the battery packs is not the main issue.
  6. Oisko sisäpuolella sen verran tilaa että nuo vois vaan pultata jos tekis reiät koteloon? Ei tarviis arpoa että irtooko Muistaakseni joskus koittaessa tuota lakia tulkata, siellä oli vielä erikseen pätkä joka sanoi että "itsetasapainottuvat" vehkeet joita saa ajaa jalkakäytävällä pitää olla sellaisia että ne pysyy pystyssä paikallaan ilman kuljettajaa, mikä käytännössä tekisi kaikesta alle 3-renkaisista laittomia jalkakäytävällä (ei pyöräteillä/kevyen liikenteen väylillä), kun ei ne itsekseen paikoillaan ja pystyssä pysy. Tiedä sit miten sitä käytännössä tulkitaan.
  7. esaj

    Headache

    I must admit that I didn't watch the videos, but I did get one of those DSO-scopes (a DIY-version, with no enclosure) back before I bought the Rigol, and already have one cheapo LC-meter (Ie. "just" capacitors and inductors, no resistors/ESR like the LCR-meter, cost about 20€, a good quality basic LCR-meter is maybe 100-200€, professional-level high precision meters of course go up into thousands as usual ) and a simple semiconductor/component tester (which also can measure capacitor ESRs, although probably not very precisely). These kinds of devices aren't really bad for their price in general, and good enough for general hobbyist usage, but I have no real use for the cheap stuff anymore really, what I'd likely need next is a more "serious" high-power lab PSU (Still looking at that 150V / 1.5kW TDK-Lambda Genesys) and another scope with much lower noise-floor, the Rigol's otherwise nice, but the noise generated by the scope itself is around 800µV, which is pretty bad considering the (possible) CD-project. What I was going to do this summer was to get my electrics rebuilt in the house... Not only would a 1.5kW power supply (if used at full power) put more stress the cabling, my current setup already likely does run pretty close to the limits This room + the one next to it comes through a single feed behind a 16A fuse, and I could be using several kilowatts already here, plus the next room has a freezer and wheel charging... All the wiring and mains cabinet is original from over 30 years ago. The house across us burned down due to electric fault in the mains cabinet a couple of years ago I sent an email to 10 companies asking for an offer, 2 responded (it's been a month or more), and at least one of them fell out directly due to the price they were offering, the other came by to look at things, and promised to send an offer later, but I haven't heard back. Looks like it'll be sometime in the future then...
  8. It may make sense in special-cases, like things that are built-to-order (and usually cost some serious $$$, like >1GHz scopes, only directly from manufacturer, prices are 5- or 6-digits), people building analog audio circuits like pedals and analog synthesizers (like a friend of mine who "lives" in our garage during the summer, building cases for them ) often have to go that route (matching components and whatnot). Never really dug deep into there, as I haven't been bitten by the bug (yet ). As for resistor tolerances, 1% basic resistors cost about 1cent/piece these days when bought in lots of 100 pieces or more, so I haven't seen the need to use 5% or more tolerances, the savings are nothing. J-Fets might be different, and like said above, analog audio circuits are a special case really. For mass-produced consumer products, I doubt the manufacturers would go the way of matching components, but likely mosfets bought at the same time are from the same batch. In general, many EE's seem to frown at paralleling mosfets instead of using higher rated single mosfet, it's considered bad practice, but it can often be cheaper to use two cheaper mosfets in parallel than one better rated one, which likely is the reason the wheel's use parallel configurations instead of "proper" single device. At least Toshiba, IRF/Infineon and Nexperia have published longish application notes on problems and issues with paralleling power mosfets specifically, and at least Infineon and Nexperia suggest that optimally the mosfets should come from the same batch if parallel configurations are used, but mostly it's just a matter of handling the gate drive "correctly" to avoid parasitic oscillations. https://assets.nexperia.com/documents/application-note/AN11599.pdf https://www.infineon.com/dgdl/para.pdf?fileId=5546d462533600a401535744b4583f79 http://www.irf.com/technical-info/appnotes/an-941.pdf https://toshiba.semicon-storage.com/info/docget.jsp?did=59458 In the past, I've done exactly one device which used paralleled mosfets (5 in parallel), but it could put out >1000A in short spikes (1-10 milliseconds, spot welder), and I followed the application notes on the basic suggestions how to ensure the turn-on and -off don't cause problems and the load is shared as equally as possible (gate resistor with anti-parallel diode to speed up the turn-off, ferrite beads to increase high frequency impedance and kill off parasitic oscillations). All the mosfets were ordered at the same time, so likely the same batch. For anything "critical", I wouldn't order the parts from Aliexpress or such, you can never know whether you get some other mosfets with changed "stamping" on the package, clones, used devices removed from a broken device, factory rejects (don't fulfill the datasheet specs) or the genuine thing. If the capacitors are discharged, you may or may not get sparks, it's all down to how good a connection you do (and how fast). The initial current when the capacitors sit at (or near) 0V, and a battery pack with 60...80...100V or whatever is connected, can be huge. If the connectors are touching "badly" (only slightly), the highest resistance point is at that connection, yet the resistance may still be low enough to allow substantial current (like 100...1000+A in a very, very short spike). Most of the voltage gets dropped over the highest resistance in the circuit, meaning most of the power is dissipated there, and the short-lived but very high power dissipations there (think 10kW or such, the above 1000+A spot welder used 12V and could go above 10kW peak when the "short circuit" starts) is enough to melt small amounts of metal (that's the sparks you see, tiny molten bits of copper or whatever the connector's made of). The battery BMS's have (or at least should have) short-circuit protections, but they still take few milliseconds to react, which is too slow to prevent this. The "anti-spark" XT90's have a clever internal structure where the connection is first made over a resistor that limits the current, but is low enough to allow the capacitors to charge before the final very low resistance connection is made, and at no point the current spikes into really high numbers. Been there, done that... Luckily no other damage than a destroyed probe and XT60-connector.
  9. Don't look too much at the wattage/current absolute maximum -numbers given on the beginning of the datasheets, they're theoretical at best and cannot be attained in reality (at least not without very efficient cooling, like pumped liquid with large external radiator or liquid nitrogen or something, and even maybe not then ). To use IRFP4468 as an example, here's how they arrive at that 520W number: Starting point: Ambient temperature at room temperature (usually 25C in datasheets). Case (package) at ambient temperature. Junction-to-case thermal resistance is 0.29C/W (ie. the junction temperature goes up 0.29 celcius above the case temperature per watt of power dissipation). Maximum junction temperature is 175C. So, assuming the case would magically stay at the same temperature regardless of junction temperature and the heat coming out from the junction wouldn't warm its surroundings, so-called "infinite heatsink-model" where the case temperature never raises, you could theoretically dissipate: Maximum power = (Maximum junction temperature - Case temperature) / (Junction-to-case thermal resistance) (175C - 25C) / 0.29C/W = 517.241W Round that up to 520W for good marketing material. The device would be working right at the theoretical maximum junction temperature limit (175C), and any external heating, or the case heating up, it would likely die. If it can even survive up to (exactly) 175C junction temperature. The thermal resistance between the case and ambient (case-to-sink + sink-to-ambient -thermal resistances) would have to be 0C/W. If the board's getting hot enough to melt solder, the ambient, sink and case temperatures are something totally different than 25C Similar "magic" is often used in the calculating the theoretical maximum currents, so you can forget about things like "single mosfet could handle the entire current, it's rated at xxxA!" if just looking at the absolute maximums. I don't know how the peak pulse currents are calculated, heat doesn't move at infinite or "even" light-speed, so with a large current spike, the heat doesn't even have time to conduct away from the junction... They're not outright lying, and all the manufacturers use similar techniques to calculate the maximums (so they look as "good" as the competition), but you can never get to those numbers in the wheel, maybe on a lab bench with very heavy cooling, for a short while. Also, typically the mosfet legs will melt before such figures are reached continuously or in average anyway (that's the "package limited" -value), pulsed they can probably handle high currents for a while, at least if allowed to cool in between. If you compare the Rds_on -resistances, the list looks like: HY5012: 2.9mOhm typical, 3.6mOhm max IXFX300N20X3: No typical given, 4mOhm max IRFP4468: 2.0mOhm typical, 2.6mOhm max IXFH320N10T2: No typical given, 3.5mOhm max You'd want as low as possible resistance, so as little as possible of power is "wasted" in the mosfet, heating it up. Another thing to note here is that the IRFP4468 and IXFH320N10T2 are a 100V mosfets, whereas HY5012 is 125V and IXFX300N... is 200V Vds max. Without knowing how high voltage spikes the motor can put out, going lower than the original might be risky. Also, I'm a little bit suspicious of the HY5012-numbers, IXYS and especially International Rectifier (nowadays owned by/part of Infineon) are big brand names in mosfets with long history and breakthroughs in manufacturing technology, a little hard to believe that some never-heard Chinese manufacturer (Hooyi?) is producing mosfets with better specs. I'd go about selecting in this order: Original (no risk of issues with wrong ratings, since it has worked in the past... well, until it blew ), or if not available/not an option, same or higher Vds_max -voltage as original with same or lower Rds_on -resistance as original. Obviously the same package (TO-247 in this case). Preferably same'ish gate charge, although someone with an EE -background once said that it likely doesn't matter that much in wheels, since the wheels use (relatively) low switching frequencies and powerful gate drives, but if the gate charging and discharging has been more or less "optimized" for some gate charge (it probably isn't, at least with GW), the gates in the new mosfets might charge/discharge too quickly (lower gate charge, possibly causing gate ringing) or too slowly (higher gate charge, excessive switching losses passing through the linear region on turn-on/turn-off). But probably it's not an issue in reality here. Safest bet is to get a whole new board, followed by replacing the mosfet with the same model as the original, followed by using another type of mosfet. Like Marty said above, there is a risk that something else has been damaged along with the failing mosfet, so it's not certain the board will work flawlessly just by replacing the fried fet, although people have replaced fets before and had the wheel working again without an issue.
  10. Oh yes, motorcycle- or traffic cops might be more "educated" on the subject. And if they can get a radar reading of your speed, and it's clearly above 25km/h on level ground (or even uphill), then things might get different...
  11. I've bashed Gotway (without ever owning one ) in the past, but this debacle isn't exactly anything we haven't seen before. Granted, GW is probably the only manufacturer that more seriously pushes the envelope (KS and Ninebot have been catching up though), but at what cost? Failure rates of EUCs are higher than most types of vehicles, but subjectively it seems GW's still leading the pack there (too)... KS has been more conservative with their designs, but they've also had issues with newer wheels (at least the lock-up and shaking issues with the newer KS18's), after a good run with the KS16-line, which had the lowest number of warranty issues a year or two back on the reseller statistics published somewhere in the forums, but those are probably outdated. And even then, the warranty-repair rate of over 1% of KS16S wasn't exactly stellar (that's still more than 1 in 100). CGI lately got a KS14 replacement board from them that had the battery connector installed in reverse (wtf?) and fried the charger and the new board. Ninebot seems to concentrate more on the (likely far more lucrative) e-scooter -business, after Z-series hasn't exactly been shining in reliability, although it seems to be mostly issues with the battery drain, and there have been rumors here that they might be leaving the EUC-business all together. InMotion had really serious battery issues with the V10-line (don't recall the model, V10F or something?), that could catch fire by itself if moisture creeped into the battery packs. GW, Ninebot, KS and InMotion seem to be pretty much the big 4 nowadays. On the smaller brands, after a couple of years of hiatus, Rockwheel released GT16, but the V1 had lots of problems, don't know how the newer V2 has fared, better I assume, after GR16 I haven't really followed up on them (and that was already released before I got here... 2014?). Supposedly they do have a new model (GT16S / Iron) coming out, but it hasn't been released yet to my knowledge. I haven't followed up on IPS, but looks like they're more or less dead in the water, I5/S5 were the latest models? IPS has had a relatively good track record on reliability, as far as I remember, but they've never made a real high-performance wheel. Maybe they're just cooking up a new model, and taking their time, which probably is not a bad idea. That's pretty much all the "bigger" manufacturers that come to mind, even though EUCs are relatively new (Solowheel early this decade, 2012 or something, the Chinese manufactures emerging somewhere around 2014-2015?) a lot of manufacturers have already disappeared entirely from the scene. We may be in something like the "Model-T -era", but I'd expect better reliability from these things, especially at the current speeds, considering how much more worse a failure can be if you're travelling at something like 30mph or such. Some people have suggested that the manufacturers should follow something more like the automotive / aerospace industry standards, but likely fail to understand the costs of such (a friend of mine, who's a helicopter mechanic, once told me that a single specially built bolt to a chopper can cost >60€ per piece, more complex parts can easily be thousands) and especially on cars, the economies of scale. Some car models are built and sold in the amounts of hundreds of thousands yearly, and in total there are millions of cars built and sold yearly, the entire EUC business is a small fraction compared to that. The requirements are much stricter, and still there are issues with them at times (albeit usually much "milder"). The cycles are longer, and more time is taken to test things thoroughly. While the software is complex, I know that the piece of software I wrote a couple of months ago won't be in real vehicles (not cars, btw) until about two years from now. Before that, it has gone through a lot of testing, and possible bugs should have been found and ironed out (hopefully... ) But maybe a lot of it comes down to us "consumers". We want (demand) more speed, more features and more power on the wheels, and we want it now, not year or two down the road. This puts pressure on the manufacturers to come up with new, faster, bigger and more powerful models, while making it look "sexy" on the outside or something and hasten the development cycle, maybe even skip proper testing, before putting the new model on the market. Unfortunately, this seems to be a somewhat general trend with many products worldwide. I started working professionally in the software-business in the latter half of last decade, when the "agile development" was picking up pace, and while it can produce good software when done "right", it feels like a lot of companies just use it as an excuse to push out their product earlier, develop things in shorter cycles and do the beta-testing on the users. I've heard the term "time to market" a bit too often. And "software can always be updated afterwards and bugs fixed, since nowadays everyone has Internet and everything's online" At times, it seems this is what's happening with the wheels also. Unfortunately, a hardware or software bug there can have much more dire consequences than with (non-critical) products. Slowing down the pace might not be a bad idea. More careful design and planning of the entire assembly procedure, proper testing cycles etc. would likely probably bring down the failure rate a lot. But that means that the buyers will need to wait longer, and many people seem to be "conditioned" to always get the newest and latest on everything. Why buy the model that has been proven over a year or two, when the new, shinier and "better" model is available very soon? The marketing on most things seems to be feeding this, you just have to get that new phone model, the 1 or 2 year old one you have isn't that good anymore I was reading the hardware product design article-series someone linked to a while ago, and while it had good points, it ended with the "planning the product lifetime" or something along those lines, which basically stated that when you're finishing up your current model, most products have a lifecycle of 18 months(!) and you should start planning how to fade away the old model and how to get people to buy your next, better, shinier product... Don't know how much of any real use my drunken ramblings are, likely none, but that's my 2 cents anyway...
  12. If the wheel's being used all the time, likely it won't sit at high charge for a long while, so mostly the degradation likely comes just from the general cycling of the battery. The degradation in storage seems to mostly occur over longer time, depending on temperature (who'd really keep their wheel at 40C or 60C = 100-140F? ) and varies with state of charge: Temperature Lead acid at full charge Nickel-based at any charge Lithium-ion (Li-cobalt) 40% charge 100% charge 0°C 25°C 40°C 60°C 97% 90% 62% 38% (after 6 months) 99% 97% 95% 70% 98% 96% 85% 75% 94% 80% 65% 60% (after 3 months) Table 2: Estimated recoverable capacity when storing a battery for one year. Elevated temperature hastens permanent capacity loss. Depending on battery type, lithium-ion is also sensitive to charge levels. ( https://batteryuniversity.com/learn/article/how_to_store_batteries ) Apart from the early "generics" (14" Airwheel X3-copies that were rampant back in 2015) and a few less known manufacturers that disappeared from the markets years ago (F-wheels used to have LiPo-packs?), I don't think any wheels used anything but big brand-name cells. Smaller capacity, but higher discharge chemistries have been used in some Gotways, I think. In the end, the voltage's going to cause a faceplant if you go fast enough (the motor back-EMF raises to the battery level => no current, no torque).
  13. An EE once told me that usually handling an assembled board shouldn't be an issue, since there are resistances and capacitances etc. on the boards that can sink the energy from the usual static discharges. Mostly it comes down to handling individual components during assembly, there the danger is much higher, and the worse kind are the "walking dead/walking wounded"-components, ie. such that have been damaged by ESD, but that don't "show" it right away, instead they'll fail hours or days of use later, but it seems quite rare. Someone here in the forums (pico?) mentioned that they used to do a "burn-in" running the devices 48 hours at full load / maximum stress to see that things work like they should, but I guess it's rare with EUCs. KS does do some sort of a running test (shown in their videos where the wheels are being run on treadmills of sorts), but no idea how long or if each and every wheel goes through this. Some components are more sensitive to ESD-spikes than others (mosfet gates come to mind, it doesn't take that much to break the insulation between the gate and the channel).
  14. The world has changed due to a lot of things being digital nowadays... In software-world, all kinds of copy protections have been used for a good while, from the old "input the word from page X paragraph Y of the game manual" to more complex schemes, like dongles that need to be connected to the computer to use the software and to on-line check of license validity and tying it to keys calculated from hardware things like network MAC-addresses these days and whatnot. Probably all of these will get broken in a matter of days or weeks after software release. I get that people/companies making the product or producing the "content", be it software, literature, movies or whatever should get compensated for their work, but it seems just weird that once you pay for a book or movie, you'd have to jump through hoops to make sure you can read/watch it again X years or decades from now should you want to. It's like selling the same content that you've already paid for again. I don't get reimbursed when someone uses a software I made years ago (granted that I was just an employee back then and the rights belong to the company I worked for at the time, but you get the idea) Nowadays, a lot of things seem to be selled based on some sort of "subscription". It's not a completely bad thing, I love Netflix for example, back when I was a poor student a long time ago, I dreamed someone would make a monthly-paid service where I could watch movies and series, and that's exactly what it does, and does well. But should they cancel my subscription for some reason, or once a series or movie gets removed from the catalogue, I can't access it. And I don't "own" any of the content there, but have a "license" to watch it. I guess it works better for movies and series than books. Slightly off tangent, the IOT-craze on the other hand seems just bizarre to me, I can see the use in industrial-settings and such (remote monitoring/control, fleet tracking, handling huge amounts of remote sensor data and such), but at least most of the stuff for normal consumers is not something they really need, if not downright dangerous, like remotely activated kettles or other high power devices that could cause a fire. And do you really need to do that remotely? A lot of these seems to be tied to some sort of subscription, rather than that you'd buy the device and "own" and use it for as long as you like (or as long as it works? ). The stupidest idea someone was trying to sell me some number of years back on the IOT-things was "smart thermostats" for the house. The idea was that they sell you the thermostats for relatively low price/margin, but you have no control over them manually. Instead, you need to login to their web-service to adjust the thermostats, and you pay monthly for the service. I then asked the guy what happens if there's no network? "The thermostats will automatically adjust to 21C". What if your company goes bankrupt? "I guess the thermostats will stay at 21C forever then". Considering that I really rarely have any need to adjust them, and can do so just by going to thermostat and doing the adjustment manually, I can't get who would really need these in a normal household... You have to be pretty lazy to need something like that, or you're obsessed with controlling the temperature room-by-room Maybe if you have something like a cottage out in the woods that you visit seldomly, and want to set the heating up while heading there, so it's nice and warm by the time you get there, but otherwise I don't get it. On a quick Google search, I couldn't find the company anymore. But there are other companies with "smart thermostats", and they do have manual control...
  15. Do you leave it charging (sometimes) until the current has dropped to zero? Did you measure the output voltage with a multimeter or just by the display on the charger? The displays look like the cheapo chinese voltage/current meters that aren't exactly precise, or at least drift over time. I have Firewheel charger that outputs around 67.6V, I initially assumed it was to overcome a reverse protection diode in the BMS, but later on learned that probably the reverse protection is done with mosfets ("ideal diode"), and it's just off... nevertheless, I've used it at times to fully charge the battery (current dropping to near zero) to make sure the cells get balanced. Never had issues with overcharging even with slightly too high voltage, probably the balancing circuits in the BMSs can handle the slight overvoltage. If you leave the "dumb charger" on until the current drops to zero, do you reach full 67.2V (or thereabouts)? If looking just at the wheel app (requiring the wheel to be turned on), it might show 67.1V just because of the current required to run the MCU & other things on the mainboard, and keep the wheel balanced on its own. Checking that the voltage is correct with a multimeter and then allowing the charge to run until (near) zero current should allow the batteries to charge all the way to full voltage. I start my rides with less than 100m of going straight and then a more steep downhill for a couple of hundred of meters, followed by more gradual decline for maybe about half a kilometer. Never had the wheels warn me of overcharge there, although it might push the voltage somewhat above 67.2V. I've done this for years, never had to do any really hard braking though. Charging the batteries all the way every now and then (say, every 10-20 charges for example?) is a good practice to make sure that the cells get balanced. For longevity, charging to less than full (most of the time) is good, also if you need to store the wheels for a longer time (I do, for about 6-7 months every year over the winter), discharge them down to something like 30-40%. If stored for longer while at full charge, the batteries lose maximum capacity much faster. If stored "empty" for a longer while and the voltage drops too much, internal short circuits will ruin the cells. Last winter, the wheels (at that time, I still had both the KS16S & KS16B) were both run down to 3 leds before putting them to "sleep", I didn't even bother checking the wheels or voltages over the winter. The voltage drop over about 7 months wasn't enough to drop a single led from the side panels (both KS16's set to show the battery state with the side leds when not moving) Ninebot Z-series is different, the vampire drain requires the wheels to be charged monthly or so.
  16. I replaced the caps for the charger today, and all I can say is that this charger ("Gojusin", apparently the manufacturer) isn't "exactly high quality". Removing the old solder from the (remaining) capacitor legs, I actually lifted of all of the capacitor pads and some of the traces! All of them! I haven't had a pad lift off since I was using some very cheap dot-matrix boards I bought from Aliexpress (and maybe too much heat) years ago. I had to bend the capacitor legs to nearest component legs on the same trace to make sure that a good enough contact is made... Many of the traces are "topped off" with solder, probably because the copper (assuming it even is copper) layer seems to be really, really thin, maybe to prevent the traces from burning out just by normal operating current. I don't know how it's attached to the substrate, by super glue? Now I kind of wish I hadn't used so "good" caps on this (120uF / 450V long life United Chemi-Con, about 3€ per piece), they'll probably outlast the charger
  17. Sounds like the fast charger output voltage is too low. I adjusted a friends' fast charger (bought from Wheelgo, Jason McNeil's UK-company before he moved to US) a couple of years back because the output voltage had dropped to too low, it was around 66V or even a little bit below, don't remember exactly anymore. It's normal for the battery voltage to drop a bit after removing the charger, to get "full" 67.2V after removing the charger, you have to wait until the charging current drops to near zero (which can take hours after the charger light already turns green) and all the cells have charged to full. Even the relatively small current required to run the electronics and keep the wheel upright at standstill (when it's own) can drop the voltage somewhat. With KS's and other wheels having reverse protection for the batteries, you can't measure the battery voltage directly from the charge port (and Charge Doctor won't turn on when plugged to the wheel without the charger plugged to power outlet), what you see during the charging in CD or similar, or in the charger itself, if it has voltage display, is the charger output voltage, which is higher than the actual battery voltage as long as current is flowing (once they reach the same voltage, the current has dropped to 0).
  18. You'd think so, but... Year or two ago (I've actually forgotten whether it was the summer of 2018 or 2017) I was riding around with a childhood friend visiting from another town, and we went to her big sister's house for a barbeque. Turns out, the big sis's husband is actually a local police officer. Over the course of the evening I talked with him about the legal situation about the wheels (they're legal up to 25km/h / 1kW, but KS16S already breaks both restrictions), and in summary, what he said was that A "normal" police officer won't know about those details, if they even know whether they're legal or not They're not usually interested in the wheels (at least in professional sense), as long as you behave and don't do anything crazy / endanger the public, they've got better things to do Even if they know the limitations otherwise, they won't know which model goes how fast or has how much nominal power They won't be able to test ride the wheel to see how fast it can go, and they can't force you to do that Finally, they're reluctant to give out tickets without a very good reason, because if they can't cite the exact laws and how they were broken, and there are some technicalities, the person receiving the ticket can "challenge" (well, that's not the right word exactly, but couldn't think of a better one right now) the ticket in court, which means they have to show up in court to give a testimony So here, your "basic" normal police officer is unlikely to stop you or give you a ticket, unless you're riding recklessly or certainly going much faster than 25km/h (think of something like riding 40+km/h or causing possible danger in sidewalks or crowds or something) and they can prove that.
  19. Yeah, that seems to be the nature of "digital" things nowadays, when you buy something, you don't own it, what you own (or have) is a license granting you the rights to use said product (with certain restrictions). I don't have a Kindle or such, but I wouldn't be too surprised if at least some of those services won't let you read the books you "bought" unless they can (at least every few days or something) check that your license is still "valid". Same goes for a lot of software.
  20. The low-side mosfet stopping the charging seems to be an IRF540, which has an absolute maximum Vds of 100V, so at least that would need to be changed to a higher voltage variety. Similar looking current/voltage-meters as what is used in Charge Doctor come in 100V and 200V -varieties, I don't know which hobby16 used (if it even is the same), but it's possible the meter cannot handle 100V and change, if it's the absolute maximum rating, but I don't know. Usually the absolute maximums are something to stay away from, ie. I wouldn't use a 100V max Vds mosfet for 100V input voltage, but maybe up to 90-95V max to leave some "headroom". Running something right at the limit at least drops the lifetime of the component, if not cause an outright failure. Still, it's also possible that the 100V & 200V -models are the same device, but just use a different voltage divider and software-values for measuring the voltage, and the circuitry otherwise works just fine with higher voltages (might need a protection diode in the measurement divider to prevent overvoltage at the MCU ADC-input). It could work, but the risk is that you'll burn your CD and/or damage your charger if it gets shorted over the failing CD. I don't want to test on mine I started polling for interest in a CD-like device some time ago, currently looking into going to open-source hardware/software-route, and maybe selling kits/pre-built devices at some point, but unfortunately the price (even just for the components + board + encasing, if building by yourself) is going to be much higher than the original CD. The current idea is to support voltages up to (around) 105V, so it can work with the 24S-wheels, OLED-display, BT-communication so a mobile app could be used and separate adapter-cables, so the same device can be used with wheels having different connectors.
  21. Yeah, I figured early that fixed connectors would become an issue, seeing that many people have more than one wheel, and quite often they may have completely different connectors, even if the voltages are the same. XT60 was just used as an example to explain the idea, I don't like it because it's "too tight", yanking the connectors off from each other can get difficult, and if the other end is board mounted, in the worst case you end up ripping the connector off the board before they loosen Molex might have some good candidates, for example Mini-Fit series has connectors where the connectors are locking (won't come off by accident, easier to detach than XT60) and the pins are rated up to 13A, and if need be, multiple pins could be used for both positive and negative (ground). I don't know what's going on with hobby16, I haven't tried to contact him. Maybe he got tired of making Charge Doctors. I'll try to keep the component costs down, but unfortunately, the "more special" components (meaning mostly the >100V switcher controller) seem to be quite pricey in low quantities, also the nRF54-SoC is about three times the price a non-BT device would need (+ it needs an antenna, but on a quick glance they're relatively cheap). Also, if the device is to end up in the hands of others, I'd rather spend more money on quality components and proper protections than "cut corners" just to keep the price down. Still can't say how much it will cost in total, as there's lots of things that are still in the air regarding the overall design, and I don't know what Seba's going to cook up, he might come up with a superior design. I was calculating the characteristics and external component values for the LT3638 last night, and it occurred to me, that while having the "support" for up to 24S (around 101V) -wheels is a good thing, not that many people have those really, most are 84V or below. Spending a bit more board space, it would be possible to add another (cheaper, but still good) lower voltage SMPS controller to the board, so that a <100V model could be made with the same board and save a bit in the component costs. Of course only one or the other controller (<100V or the LT3638) would be installed on the board, but people who don't want to pay extra for the wider input voltage range could build it/get it slightly cheaper.
  22. esaj

    18L Shakes!!!

    There are so-called "sensorless" motor driving algorithms that rely on the voltage and phase current measurements to timing the motor drive, but I don't know much about the details of motor control algorithms or if such is used in any wheel: All I remember is reading somewhere is that it's pretty hard to get right because of the precise timing requirements on the voltage measurement, and as usual with complex motor algorithms, the math gets somewhat hairy fast... Might be able to test by doing what you suggest, using some sort of a jig to hold the tire off the ground with the mainboard visible, speed it up to high speed by gently tilting and then yanking off the hall-sensor -connector, but I doubt anyone really wants to try that For science?
  23. Agree, probably the best bet is checking that the battery charges to full voltage (as said above), and if it doesn't, figuring out whether it's a faulty battery or a charger with too low voltage. Unfortunately at least with most wheels checking the battery pack health further isn't easy for a layman (even trying to decipher whether the voltage values are normal or not may not be easy), and not many people will want to unwrap the pack and start measuring individual cell voltages. Even less exchanging faulty cells. There should at least be some type of internal monitoring that can warn the user (through app or blinking lights or warning sound or whatever) if the battery voltage/behavior seems abnormal (ie. very little or no current running through the motor but the voltage drops faster than expected, tracking charge and discharge to detect energy "disappearing" from the pack or something...). Hopefully this will change in the future.
  24. I don't know about InMotion specifically, but most stock chargers are 2A or thereabouts. The maximum charge and discharge currents are usually rated with so-called "C-rates" ( https://batteryuniversity.com/learn/article/what_is_the_c_rate ), for most lithium-chemistries, the recommended charging rate is 0.5 (2 hour charge) or 1C (1 hour charge). Faster charging stresses the cells and drops their lifespan. With multiple packs in parallel, more current could be used than 2A safely, for example, the 16S4P / 840Wh uses 3.5Ah cells, 4 in parallel, meaning 14 amphours. Theoretically, the paralleled packs could be charged with even up to 3.5A per pack (14A total), but the charge port, wiring and possibly components in BMS might overheat / melt at that high current, so not really recommended. For wheels with big batteries, 5A should usually still be safe. The wheels with dual-ports can maybe be charged with even higher currents safely (10A?), but better check the wire gauges and that the BMS can handle that on the charging side, to prevent damage. Using low-side cut-off with low Rds(on) -mosfets, this could be handled, but it again pushes the prices of the components up, but not horribly. On a quick glance, for example 120V N-channels with Rds(on) around 10milliohm can be had for about 1.50€/piece, pushing 10A through one of those would "only" dissipate about 1W, which isn't difficult to deal with (for example, PowerPAK SO-8 has 20C/W maximum junction-to-ambient, 1.2C/W maximum junction-to-case, stick it to double-sided plane and stitch with thermal vias, it shouldn't warm up hardly at all, shouldn't even need much board space?). The downside is that the ADC-resolution is limited, so using higher current range (I was originally thinking up to 5A), some resolution is lost. But for 12-bit ADC (assuming the entire voltage range could be used, but depending on commonly available sense-resistor values and fixed gain current sense amplifier, it might not be), this would mean 10A / 4096 = 0,00244... A or about 2.5mA per bit, so not that bad Actual accuracy will likely not be that high, although oversampling might be able to keep the resolution high. Also, the limiting factor may be the wheel wiring, charge port or BMS, as mentioned above. If the two charger were separately connected to the device, they could be measured separately (ie. "only" 5A per channel), keeping the resolution higher, but this would make the device bigger and maybe somewhat more complex, at least more expensive, as certain parts would be duplicated. I'm not sure if @Seba still started to design anything, he might have his hands pretty full with other stuff. Unless he's going to do the design, I'll try to get started soon...ish, likely I'll go with one of the LTC363x buck-controllers mentioned before, at least for the time being, unfortunately they're a bit expensive (likely more than 10€ with the inductors and capacitors, possibly a diode if using the non-synchronous version). Things can always be of course changed later on as needed. There are likely parts where I'll need help, like for example I have no idea how the BT-antenna dimensions are calculated, if using a PCB-antenna, or of impedance matching if using chip-antenna, although probably there are good resources available for both, and it's not the first thing I'm going to be tackling anyway.
×
×
  • Create New...