|
ante posted:Always order an extra ^^^^^^ A good repairman anticipates failure.
|
#
?
Apr 10, 2023 02:21
|
|
|
|
| # ? Jun 13, 2024 05:15 |
|
|
kid sinister posted:^^^^^^ When the repairman is cruft, anticipate six failures.
|
|
#
?
Apr 10, 2023 02:43
|
|
|
Ferrite vs metal inductors for switching supplies. Is there some frequency rule of thumb for switching supplies? I guess at 2 MHz, I'm now solidly in ferrite-land, is that right?
|
|
#
?
Apr 10, 2023 05:34
|
|
|
ryanrs posted:Ferrite vs metal inductors for switching supplies. Probably, if you want best efficiency that is. 2MHz is quite fast, and ferrites typically have lower core loss. But ferrite and metal/composite/powder are broad categories, and there's a lot of other factors to worry about. Linear Tech/ADI usually recommend specific inductors in their documentation. Also look at their reference/eval board designs. ANIME AKBAR fucked around with this message at 06:55 on Apr 10, 2023 |
|
#
?
Apr 10, 2023 06:53
|
|
|
Stack Machine posted:It's no sweat. Those packages have wettable flanks so, except for the exposed pad, you can see the little blob of solder sticking to the side of the pin in the microscope to verify it's soldered down. I've swapped them with just hot air, wire solder, and solder wick before and it wasn't bad, and I'm saying this as someone who solders out of necessity a lot but never took any time to learn how to do it properly. ryanrs posted:Ferrite vs metal inductors for switching supplies. As AA said, follow the datasheet. They usually give you explicit layouts, including values. Also, ferrite beads fall into the category of "if you don't know what you're doing, it will probably come out worse than leaving the part out" Dominoes fucked around with this message at 21:21 on Apr 12, 2023 |
|
#
?
Apr 12, 2023 21:18
|
|
|
I thought you were supposed to sprinkle beads over your circuit to ward off EMI, like ferrite holy water?
|
|
#
?
Apr 13, 2023 01:12
|
|
|
ryanrs posted:Ok, sounds good. My main concern was stacking too many ‘first time’ experiments onto this one board. But it sounds like it’s not that bad.
|
|
#
?
Apr 13, 2023 01:34
|
|
|
Forgot to mention! For extra fun, the board is double sided: I feel like those ceramic caps really, really want to be on the back of the board, close to the switcher module pins. I guess I should add thermal reliefs on the backside copper fills, since I will be reflowing with hot air only (and maybe a soldering iron)? I did not put thermal reliefs on top-side SMT components because they will be reflowed on a skillet. All through hole components have relief, except the switcher module which runs 30A through a single pin. PCB stackup is 2.4mm thick, with 2 oz outer layers and 1.5 oz inner. Rework will be way more annoying than on a normal pcb.
|
|
#
?
Apr 13, 2023 02:11
|
|
|
ryanrs posted:I thought you were supposed to sprinkle beads over your circuit to ward off EMI, like ferrite holy water? Not a bad approach really. I've been taught to put ferrites on all power rails and signals that are leaving/entering the board, and if you care about inter-circuit EMI/EMC, each subcircuit. Inter-circuit EMI/EMC matters a lot if you're doing mixed analogue/digital stuff, or anything with a radio. it can also be easier to sort out a particularly noisy subcircuit before the noise gets somewhere harder to fix - like a cable leaving the enclosure. The recommendation I've heard is generally pi filters are better for high impedance loads (EG signals, low power subcircuits) and T filters are better for low impedance loads (subcircuits) - so I just plonk T filters on power rails, pi filters on signals I think are going to matter (generally external connections), and just ferrites or 0R links by themselves on signals I don't think will be a problem* (generally internal connections). High speed poo poo gets treated on a case by case basis, but given most high speed stuff is differential, I usually go for a common mode choke explicitly designed for the high speed signal in question. If you're really doing things properly, you line all your filters up neatly so you can smack a can over the top - though trainining I did recently indicated a lot of the time if you design for a CAN you end up not needing one. The only nasty thing I've been warned about with ferrites is there will be certain frequencies where they're inductive, depending on the ferrite, and you can acidentally hit a resonance - then they do more harm than good. I've never seen this happen myself though, and if it did the fix would be as simple as replacing the ferrite. The other nice thing about ferrites everywhere is they provide an easy way to isolate subcircuits and modify power and signal lines, which can be invaluable if you're doing a low power design (and trying to work out whats drawing what). *if I'm wrong, it's not an impossible mod to bodge the 0R link into a pi filter with a bit of scalpel work and swearing
|
|
#
?
Apr 13, 2023 09:24
|
|
|
ryanrs posted:I thought you were supposed to sprinkle beads over your circuit to ward off EMI, like ferrite holy water? Nah, you're thinking of bypass capacitors. Ferrite beads/supply chokes are more like weed killer. Think first. They're something I only use if it's explicitly recommended by an app note, or I have a very clear and testable purpose in mind. Splode posted:Not a bad approach really. The real trap is when people start throwing down board-level common mode chokes, or putting chokes on GND. There are cases where doing so is helpful, but they are very specific and contextual, and it's very easy to change a perfectly valid scheme into a clusterfuck by adding just one more choke or cable. If ferrite beads are weed killer, this poo poo is more like chemotherapy. Do not gently caress around. In a recent job it was very clear that the other EEs thought common mode chokes were pixie dust which you could sprinkle around like ceramic bypass capacitors. Any one of them could have been justifiable by itself, when looking just at a certain bit of the overall design. But together it was a complete shitshow where injecting current/voltage at any point would cause various signal levels on multiple boards to bounce around. My favorite example was a big common mode choke on a power supply. But the circuits on both sides were connected directly to chassis ground, effectively shorting out the entire choke, making it completely useless. The chassis connections weren't obvious when looking at individual schematics. Even after pointing out the issue, they insisted the choke had to stay (but were never able to demonstrate why). It just made them feel better. One lesson I learned early on is, if you use a common mode choke between two different boards/systems, then every other connection between the two must either: 1. Pass through the same common mode choke (as in, be coupled to the same core, not put through another choke in parallel to the first) 2. Be galvanically isolated 3. Be fully differential (with a high common mode impedance) ANIME AKBAR fucked around with this message at 13:31 on Apr 13, 2023 |
|
#
?
Apr 13, 2023 13:24
|
|
|
God that sounds like a nightmare. I did some EMC training recently and the guy opened with "If you can, just use fibre optics to communicate between devices, and then you don't have to worry about any of this poo poo" before launching into the details
|
|
#
?
Apr 14, 2023 01:06
|
|
|
ANIME AKBAR posted:
This would drive me up the loving wall
|
|
#
?
Apr 14, 2023 01:15
|
|
|
Kinda amazed my own designs have had more thought put into that poo poo than people paid to do this.
|
|
#
?
Apr 14, 2023 01:39
|
|
|
    Board is almost done!  Bottom-side capacitor thermal reliefs. Some issues remain, like what is going on with the U9 QFN paste stencil? The aperture looks pretty under-pasted, especially with 6 vias wicking away solder. I downloaded the footprint from SnapEDA, so I better double-check it. OTOH, solder bridges under the chip would be a real pain in the rear end, so maybe a sparse paste pattern is intentional? 
|
|
#
?
Apr 14, 2023 02:56
|
|
|
Out of curiosity, what does it cost to have a few of those boards made, all-in? Seems to be similar complexity to those DisplayPort adapter boards you saw me posting, so now I'm wondering if I can roll my own instead of working with these AliExpress specials. I've seen online services that let you design a pcb cheap, but are there some that will place and solder all the components too, without a  premium?
|
|
#
?
Apr 14, 2023 03:11
|
|
|
That particular board will cost about $320 for the pcbs (PCBWay, 5 boards min order) and $200 in parts from Digi-key. But this is an expensive PCB (thicker fiberglass core, extra thick copper) with fancy, expensive chips (esp. Analog Devices). If I wanted the pcb factory to also do the soldering, that would add maybe another $100 or so. For a simple, little 2-layer board with a couple cheap chips, you could get 10 copies manufactured and tested for maybe under $100 (<$10/ea). I'm not sure where the floor is, but it's something like that when you're talking 5 or 10 units. If you want 100 or 1,000 the per-unit cost goes way down. These prices are for PCBWay and JLC (cheaper). Both are in China and produce good quality boards, have good English support, etc. e: I have some posts in my AI thread with detailed cost discussion. https://forums.somethingawful.com/showthread.php?threadid=3961355&userid=0&perpage=40&pagenumber=23#post526879563 https://forums.somethingawful.com/showthread.php?threadid=3961355&userid=0&perpage=40&pagenumber=24#post527070332 https://forums.somethingawful.com/showthread.php?threadid=3961355&userid=0&perpage=40&pagenumber=25#post527890652 https://forums.somethingawful.com/showthread.php?threadid=3961355&userid=0&perpage=40&pagenumber=26#post528301104 ryanrs fucked around with this message at 03:46 on Apr 14, 2023 |
|
#
?
Apr 14, 2023 03:24
|
|
|
Standard process boards are cheaper than that. Default toleranced JLPCB is $13 total for 10x 2-layer boards of that size. 10x 4-layer boards of that size are $45 total. Most of that is still setup cost, additional boards are still ~$1
|
|
#
?
Apr 14, 2023 03:40
|
|
|
For 1 or 2 boards, the real tradeoff is do you want to solder them, or do you want to write up the specs and instructions and and BOM spreadsheets to have someone else solder it?
|
|
#
?
Apr 14, 2023 03:52
|
|
|
Zero VGS posted:I've seen online services that let you design a pcb cheap [..] FYI, don't use some pcb manufacturer's bespoke web app to design boards. That locks you into their production (and higher prices). Design your boards in KiCad, which is free and better than some manufacturer's app, then get your boards made literally anywhere in the world.
|
|
#
?
Apr 14, 2023 05:38
|
|
|
Just make sure you double check the pinout on your footprints. Got burned recently because my teammate designed a board but didn't check that the pins on a mosfet were correct and the circuit obviously didn't work. You'd think of all things, a 3 pin mosfet would be correct, but it was wrong.
|
|
#
?
Apr 14, 2023 05:51
|
|
|
Cojawfee posted:Just make sure you double check the pinout on your footprints. Got burned recently because my teammate designed a board but didn't check that the pins on a mosfet were correct and the circuit obviously didn't work. You'd think of all things, a 3 pin mosfet would be correct, but it was wrong. Ah yes, the rare 60 degree mounting component 
|
|
#
?
Apr 14, 2023 06:06
|
|
|
SOT-23FU
|
|
#
?
Apr 14, 2023 06:13
|
|
|
Splode posted:… Don't forget their largely impossible to predict hysteresis! Took me a while to work out why a speaker sounded like absolute crap, turns out a particular Vishay 0805 ferrite has massive hysteresis, so even driving a 1 W 8 Ω speaker at low levels caused distortion so bad you could see it on an oscilloscope. Looked almost identical to crossover distortion. ANIME AKBAR posted:… Not to defend cargo cult engineering, but this sounds like how I've had automotive EMI design described? Power fed by twisted pairs, going to non isolated loads that are also chassis connected. It's expected that some DC current will return in chassis, but the idea with the choke is that any differential mode ripple current will be forced to primarily return through the twisted pair by the common mode choke (and the TP itself). Same idea as putting isolated DC/DC converters on all power inputs between boards and using differential signals, but cheaper since we know that from an EMI perspective you don't really need to keep return paths under control below some frequency. Made sense to me but I guess that approach might have some problems, I've never actually designed a system with that topology. DC/DC converters make these problems much easier to understand, since as soon as you start using them they will be the source of all future EMI problems.
|
|
#
?
Apr 14, 2023 07:14
|
|
|
ryanrs posted:FYI, don't use some pcb manufacturer's bespoke web app to design boards. That locks you into their production (and higher prices). Oh yeah I was trying out KiCad. ryanrs posted:For 1 or 2 boards, the real tradeoff is do you want to solder them, or do you want to write up the specs and instructions and and BOM spreadsheets to have someone else solder it? Well if I'm hand-soldering them I assume I'd need to order all components with leads from Mouser/Digi-Key? Is that cheaper than having a PCB shop use their SMT components (which I assume they get favorable bulk prices for) and do that computerized placing? edit: I was looking up SMT soldering stuff and ran into this video which is pretty hillarious and I assume not practical at all, but clever idea: https://www.youtube.com/watch?v=r0csHZveVvY Zero VGS fucked around with this message at 14:08 on Apr 14, 2023 |
|
#
?
Apr 14, 2023 14:04
|
|
|
Zero VGS posted:Well if I'm hand-soldering them I assume I'd need to order all components with leads from Mouser/Digi-Key? Is that cheaper than having a PCB shop use their SMT components (which I assume they get favorable bulk prices for) and do that computerized placing? quote:edit: I was looking up SMT soldering stuff and ran into this video which is pretty hillarious and I assume not practical at all, but clever idea:
|
|
#
?
Apr 14, 2023 14:41
|
|
|
longview posted:Not to defend cargo cult engineering, but this sounds like how I've had automotive EMI design described? I put together a simple LTspice sim (hopefully image attachments work). Lp1/Ls1 are the windings of a common mode choke carrying DC power from source V1 to load R3. Rp1/Rs1 are the resistances in each winding of the choke. R1 is the resistance of the chassis connection between the ends of the "cable". If R1 were open circuit, then Lp1/Ls1 carry equal and opposite currents, meaning the common mode current is zero, great. But if R1 is comparable with Rs1/Rs2, then some of the return current through Lp1 returns through R1 instead of Ls1. This means Lp1/Ls1 don't have equal/opposite current, meaning there is a DC common mode current in the choke. The reason this sucks is that common mode current magnetizes the core, and can therefore cause the choke to saturate, which means it probably won't perform its intended role very well. The real lovely part is that manufacturers basically never characterize the saturation point on common mode chokes. I've literally never seen it in a datasheet (except coupled inductors which are effectively common mode chokes but intended to be used with a magnetized core). And manufacturers tend to optimize the design to give the highest common mode impedance, which means using the highest permeability materials, but reduces saturation current (but to what, you can only guess). The few times I've asked about specs with nonzero common mode current, I'm told that it's not characterized and is not recommended. If you do poo poo like you describe for automotive systems, be prepared to discover that your chokes don't always work the way you expected. But it will depend on how much rust there is on the chassis, and how much torque is on the bolts, and the last time it rained, etc. Oh and if R1 changes quickly (because all the connections are vibrating and the chassis is twisting), have fun dealing with all the resulting transient poo poo (both common and differential mode) this will inject into this system. ANIME AKBAR fucked around with this message at 15:08 on Apr 14, 2023 |
|
#
?
Apr 14, 2023 14:51
|
|
|
For me this really sunk in during a project in which I basically made a big H-bridge motor drive (so outputs were fully differential). I wanted some common mode filtering but didn't want to use a single giant common mode choke, so I tried putting two "identical" chokes in parallel (soldered on the same board). What I found was that when the motor current was near zero the chokes worked as expected. But if I increased output current beyond a couple amps, suddenly it was like the chokes weren't there at all. But if I removed one of the chokes, then it worked well no matter how much current I put through it (until it overheated of course). Attached sim shows the problem. Each box is one of the two chokes. But there's a little extra resistance (R2) on one winding of one choke. This fucks everything up. If R2 is not there, then both chokes carry equal DM current, and both carry zero CM current. But with R2, the currents redistribute such that they now carry unequal DM currents, and they both carry nonzero CM current. That CM current causes them to saturate, and thus not provide any filtering benefit. The severity of the effect depends on the value of R2 relative to the resistance of the other windings (and other common resistances). In my case R2 must have been under a milliohm. But that was enough to cause the chokes to saturate when the output current was just 10% of their rated current. Of course, even a simulation won't show the effects of the saturation, unless you go to the trouble of actually characterizing it in the real part (the manufacturer won't help you) and then put that behavior in the spice model. ANIME AKBAR fucked around with this message at 15:05 on Apr 14, 2023 |
|
#
?
Apr 14, 2023 15:02
|
|
|
Design an actively-balanced common mode choke, with a 3rd winding (w/ higher turn ratio) to cancel the current imbalance, ha ha.
|
|
#
?
Apr 14, 2023 17:59
|
|
|
ryanrs posted:Design an actively-balanced common mode choke, with a 3rd winding (w/ higher turn ratio) to cancel the current imbalance, ha ha. IIRC the V2 missile's analogue control system used poo poo like that cuz vacuum tubes were too delicate to survive launch so sure why not.
|
|
#
?
Apr 14, 2023 18:42
|
|
|
Hobbies, Crafts, & Houses > Learning electronics: You know who else designed an actively-balanced common mode choke?
|
|
#
?
Apr 14, 2023 20:36
|
|
|
Shame Boy posted:IIRC the V2 missile's analogue control system used poo poo like that cuz vacuum tubes were too delicate to survive launch so sure why not. I'm guessing you're referring to mag amps. Very cool poo poo, too bad it's hard to get good square loop materials now. Always wanted to try building stuff with saturable reactors. I think they're still used in very high power welders, electrolysis setups, and other applications where you need to control insane DC currents. ANIME AKBAR fucked around with this message at 13:05 on Apr 15, 2023 |
|
#
?
Apr 15, 2023 13:03
|
|
|
Also in obscure power amplifier technology, the Amplidyne https://en.wikipedia.org/wiki/Amplidyne It's basically a motor-generator amplifier, getting into pretty unfamiliar territory for me. These were used for various things, one place I know they were used was for the 30 and 12 foot ship mounted tracking antennas during the Apollo missions. Very good points! I don't do automotive stuff, but this actually came up when talking grounding philosophies with one of the EMC people at work. I think it's a valid way to design in some cases, like if both ends of the link also share a solid reliable ground plane or the cables have proper shielding. Several potential issues there though, including the safety aspects of current carrying chassis connections. You're right that it's very rare to see a common mode saturation specification. I've also heard of saturation problems where EMI filters are too capacitive and this can cause choke saturation. AFAIK that one took a long time to figure out since there were about a million other things that could cause the noise problems they had.
|
|
#
?
Apr 15, 2023 14:08
|
|
|
I found some Chinese YouTube where people are designing the same eDP to Displayport adapter boards that I'm trying to buy/fix/make. Someone in this video is soldering all the SMT with nothing but a bigass soldering iron and some tweezers?!: https://www.bilibili.com/video/BV1Ut4y167yi/ That seems pretty crazy, I would think you'd at least want one of those SMT pencil-grip heat guns and some solder paste, or something. The eDP socket they're soldering is 40 pins all a fraction of a millimeter apart and they seem to be freehanding it and cleaning it up with a solder sucker between passes. It seems like someone open-sourced the design on this other site? Seems like I can't sign up to download the files without a China phone number though. Still, fascinating stuff, I haven't found any English-speaking sites with people making these boards. https://oshwhub.com/lnksy/edp-40pin
|
|
#
?
Apr 15, 2023 15:48
|
|
|
Yeah anything looks crazy when sped up 10x.
|
|
#
?
Apr 15, 2023 20:51
|
|
|
https://i.imgur.com/p4ew3u6.mp4
|
|
#
?
Apr 15, 2023 21:02
|
|
|
ANIME AKBAR posted:I'm guessing you're referring to mag amps. Very cool poo poo, too bad it's hard to get good square loop materials now. Always wanted to try building stuff with saturable reactors. I think they're still used in very high power welders, electrolysis setups, and other applications where you need to control insane DC currents. Yeah one of the things I've wanted to do for a while is play around with weird-rear end magnetic circuits like mag amps or ferrite bead logic gates or the other early solutions to "transistors haven't been invented yet / were just invented and are low power and unreliable, and vacuum tubes are way too fragile"
|
|
#
?
Apr 15, 2023 23:44
|
|
|
Also kinda wanna build something out of nuvistors one of these days, RCA's attempt to compete with these newfangled transistor things by just making... itty bitty vacuum tubes. I think they're mostly ignored these days by vacuum tube weirdos because they don't look very cool and they're metal-cased so you don't get that nice vacuum tube glow or anything.
|
|
#
?
Apr 15, 2023 23:52
|
|
|
Zero VGS posted:I found some Chinese YouTube where people are designing the same eDP to Displayport adapter boards that I'm trying to buy/fix/make. A big FFC connector can also be awful, but look like they did pretty well on that one despite not having a drag hoof. lmao I remember soldering like this. Wish I could go back in time and show my younger self what the results look like under a microscope, or under an x-ray. And maybe save myself a couple agonizing incidents of tiny molten solder/rosin bits splattering on my eyeballs. ANIME AKBAR fucked around with this message at 00:07 on Apr 16, 2023 |
|
#
?
Apr 16, 2023 00:01
|
|
|
longview posted:Also in obscure power amplifier technology, the Amplidyne https://en.wikipedia.org/wiki/Amplidyne quote:I don't do automotive stuff, but this actually came up when talking grounding philosophies with one of the EMC people at work. Fortunately, it often turns out that even if most of your chokes are saturated, they probably weren't necessary in the first place (like most of the MLCCs on your 3.3V rail), so you end up passing CISPR somthing-teen class whatever anyways. Also I should mention that the problems with saturating common mode chokes mainly applies to chokes made from winding many turns around a core. On the other hand, the ferrite tubes you often seen integrated into cables (often called ferrite beads instead of common mode chokes, but are doing the exact same thing) usually can put up with a decent amount of common mode current without saturating. ANIME AKBAR fucked around with this message at 05:28 on Apr 16, 2023 |
|
#
?
Apr 16, 2023 00:26
|
|
|
|
| # ? Jun 13, 2024 05:15 |
|
|
Shame Boy posted:Yeah one of the things I've wanted to do for a while is play around with weird-rear end magnetic circuits like mag amps or ferrite bead logic gates or the other early solutions to "transistors haven't been invented yet / were just invented and are low power and unreliable, and vacuum tubes are way too fragile" My favorite is core rope ROM. Each single core is an entire word of ROM, including the address decoder. You program it by threading the correct bit lines through each core. Since during a given read operation only one core will have current on all of its address lines, only that core gets reset and whatever set of bit lines passes through it will get pulses.
|
|
#
?
Apr 16, 2023 01:03
|
|
