#avr Logs

Mar 28 2021

03:42 AM rue_shop2: the print job from my special dos station that used to take like 10 minutes now takes 7 seconds
05:05 AM exp: rue_shop2: congrats
05:05 AM exp: i'm trying to determine how much damage I'll cause putting 100W through a 3W power resistor for 30ms
05:06 AM cehteh: likely none
05:06 AM cehteh: what kind? wire, oxide?
05:06 AM exp: the only datasheet that specifies it is Ohmite, and yeah probably oxide
05:06 AM exp: doing inrush control the laziest way, a resistor and a comparator
05:07 AM exp: just... quite a lot of inrush
05:07 AM cehteh: while i would go for 5W because they have considerably more mass which makes a lot difference when it comes to pulses
05:07 AM exp: https://www.ohmite.com/assets/docs/res_tkh.pdf pulsed power ratings don't extend past 10ms, but are for repeated duty, and with a heatsink, and also the graph isn't great
05:07 AM exp: that's the 45W option but I believe it requires a heatsink
05:08 AM exp: I'm wondering how much I can go down, and yeah it's total energy i'm struggling with
05:08 AM exp: rather than power
05:08 AM cehteh: ah oh smd
05:08 AM exp: i can THT if needed
05:08 AM exp: but one less assembly op
05:08 AM cehteh: its all about mass when it comes to short time energy disscipation
05:09 AM exp: yeah exactly, this looks semi ideal but i'd prefer something in 2512
05:09 AM exp: tbqh i might have to make some ghetto constant current loop
05:10 AM cehteh: some huge copper areas may discipiate heat, but that only works on longer timeframes
05:10 AM exp: oh the board itself is a superb heatsink that part is fine
05:11 AM exp: it handles 15-20A at 30v without any issue which is nice
05:11 AM cehteh: enough space for 2 resistors in series?
05:11 AM exp: just the lack of proper pulse rating from most manufacturers is irritating
05:11 AM cehteh: that doubles the load capacity
05:12 AM exp: it does, but I really want to get this right first time
05:12 AM cehteh: imo if applicable thats pretty ok to do
05:12 AM exp: i could also double the resistance to lower peak charge current and then switch out resistor #2 after a bit of charging, but might as well try and make a loop there
05:13 AM cehteh: ntc?
05:13 AM exp: ooh, now that's an interesting thought
05:13 AM exp: I've never used them
05:13 AM exp: but I close an external switch that shorts out this section once the voltage raises to within a couple of v of bus
05:13 AM exp: so that might actually be exceedingly elegant
05:13 AM cehteh: usually you use ntc's for such inrusch limiters
05:14 AM exp: I haven't had much experience doing this other than using ICs, so thank you for the advice cehteh much appreciated
05:14 AM cehteh: but be careful when you get pulses while the ntc may be still hot it has initial low resistance
05:14 AM exp: yeah that's fine there's current limiting upstream, but there might not be in future models, so I need to get this right
05:14 AM cehteh: also usually they dont get hot
05:15 AM cehteh: dont forget to put a fuse in there :D
05:15 AM exp: not sure what fuse options I have tbqh
05:16 AM exp: I'm using a modular supply that has a programmable current limit, so for now that's my protection
05:16 AM cehteh: might be ok too .. just make sure its protection is really working
05:16 AM cehteh: and input fuse on AC side of the supply
05:17 AM exp: yeah it's a proper supply, medical rated etc
05:17 AM exp: no messing around there
05:18 AM cehteh: looks good then
05:19 AM exp: digikey has some decent sized through holes, but no larger wattage NTCs I can see btw
05:19 AM exp: if you can make a recommendation that'd be interesting, but otherwise I'll potentially use the through hole ones
08:48 AM specing_ is now known as specing
12:28 PM rue_mohr: exp look at the duty, if the avarage is over 3W it'd be an issue
12:42 PM cehteh: thats not so simple
12:43 PM exp: i ended up deciding just to rearchitect it into a constant current charging circuit for now fwiw
12:44 PM exp: toshiba and vishay both do FETs that can survive the initial energy dump, and they use effectively less area and can use the same sense resistor i'm already using
12:44 PM exp: it also lets me more easily tune the inrush by swapping a setting resistor instead of a gigantic power resistor or a tricky to arrange NTC, although NTC is a good solution it doesn't fit ideally with my use case
12:49 PM cehteh: i didnt even followed what you are exactly doing
12:49 PM cehteh: whats the inrush current from? charging a cap?
12:50 PM exp: that and a bunch of downstream stuff that has inconsistent draw on startup
12:50 PM exp: about 10mF adjacent to the board
12:50 PM exp: unknown amounts further down the line
12:50 PM exp: at the moment I'm using the startup of a modular power supply to do the initial current limited charging
12:50 PM cehteh: can you filter it with an inductor? (without getting swings :D)
12:51 PM exp: possibly but I suspect it'd have the exact same issue as the power resistor
12:51 PM exp: I don't really have a problem with using an opamp and FET as I have a spare opamp onboard already
12:52 PM exp: so now I just have to fight with Toshiba providing encrypted PSpice only :(
12:53 PM cehteh: often enough current peaks are not that dramatic because of stray inductance and resistance on traces and components
12:53 PM cehteh: still need to be accounted for but dont worry overly much and/or make a test setup and see how it behaves in realit
12:53 PM cehteh: y
12:54 PM exp: oh I have, we're blowing right past the energy limit of our selected high side switch
12:55 PM exp: if I leave it open while I charge up the capacitance, it's quite happy in the active current limiting loop
12:55 PM exp: just there's a good many mJ to be dissipated in bringing 10mF up to 30v or so
12:58 PM cehteh: can you limit the charge current for the cap only ?
12:58 PM exp: that is more or less what i'm trying to decide on how to do
12:59 PM exp: or do you mean can i lower the current limit on the switch?
12:59 PM exp: unfortunately not on this model, the TPS27S08 or something i think offhand
01:16 PM cehteh: thats not completely trivial but there are solutions for that
01:17 PM cehteh: can you afford an relay there? ntc with relay in parallel
01:18 PM exp: well if I added an NTC I could just use a FET there as long as it was switched fast enough
01:18 PM exp: the problem with the NTCs is more mechanical space, I might end up going for it
01:18 PM cehteh: the trick is using a relay because its slow action and other characteristics (close to zero ohms) are favorable
01:19 PM cehteh: btu yes thats why i asked relay is even bigger
01:21 PM exp: yeah relay would be quite large, certainly wouldn't have to worry about a few mJ though :)
01:22 PM exp: I don't really see any major downside to just using a constant current circuit. Easy enough to enable the high side switch only after the output charges up
01:22 PM cehteh: the trick is that the relay wonrt swich much load, you stil dum the energy in a ntc
01:22 PM exp: I already have a high side 0.005R in there so I can use that as part of the loop
01:22 PM cehteh: https://tinyurl.com/ydk9prpt
01:24 PM cehteh: relay switches as soon the cap is reasonably charged, and the NTC can cool down and is ready for reuse
01:25 PM cehteh: but tricky to select a proper relay for that but such a thing is pretty common
01:25 PM exp: yeah forgive me if this is wrong but i'm a little sceptical of that
01:25 PM exp: that relay won't switch particularly well, and i'd be scared you'd weld the contacts
01:27 PM cehteh: most relays are flipover they make or make not contact, for opening you may add a snubber/varistor but for closing its pretty safe since the cap is charged already there only the normal load current not the inrush current the relay needs to switch
01:28 PM exp: well in this example if i disconnect the load it still has about 20% of the charge by the time this relay switches, and obviously that's going to vary a lot in service
01:28 PM cehteh: thats not my idea, thats pretty common circuit ... you know old amps and other things with big trafos you often hear a relay snap when you turn them on
01:28 PM exp: no i appreciate that, i just don't know if there's any significant advantage
01:29 PM exp: the relay is as large as, and costs as much as a big SOA FET anyway
01:29 PM cehteh: its has advantages and disadvantages
01:29 PM cehteh: yeah there are IC that contol a fet maybe thats better for you
01:29 PM exp: I'm no expert so forgive me if I am ignorant of some factor here
01:30 PM cehteh: imo its a very elegant application for a relay *when its done the right way* because you can as well mess it up choosing the wrong relay
01:30 PM cehteh: but that applies to fets as well
01:30 PM exp: well I have 2 spare opamp channels so I figured I'd compare the current across the power resistor against a ref and use that to set constant current, then use the second as a comparator to enable the main high side switch (and the equipment) once it's charged
01:31 PM cehteh: note that fets can have a "fails short" failure mode
01:31 PM exp: yeah good point, there's no fusing on this board but I can probably add that this revision as I'm enlargening somewhat
01:32 PM cehteh: is VCC therre and are your opamps in the stable range then?
01:32 PM exp: at 15-20A I can just add a small PCB trace and that'll work well
01:32 PM exp: I'll let you know after I see if toshiba's model works in PSpice for Ti :-)
01:32 PM exp: vishay seem to offer an unencrypted subcircuit so i'll prly just sim it with that first, they always seem to be the high end of any FET choice I make
01:32 PM cehteh: i am not so much expert either, just wanted to pointed out that old style solution which proven pretty reliable
01:33 PM exp: it's funny you mention it because elsewhere in this system we do indeed go with relays
01:33 PM exp: but I have had issues with switching them reliably due to limited coil current
01:33 PM exp: so have had to ensure they're quite well driven
01:33 PM cehteh: lol usually i suggest people not to use relays except few cases
01:33 PM cehteh: like one main relay you close before the fets and open after for reliability
01:34 PM exp: once we switch out supply that is likely to become something we do, probably an industrial contactor I expect
01:34 PM cehteh: relays are not so simple as many people think :D
01:34 PM exp: but for now we have a modular powersupply with its own limiting, hence why I can afford to redo this board and move that functionality onto it
01:35 PM cehteh: also relays want the sparks extiguished with snubber/varistor *BUT* hell breaks loose when you think you dont need that on FETS' :D
01:35 PM exp: yes there's a rather substantial unidirectional TVS on this but I frankly wouldn't mind a bit more
01:36 PM cehteh: yeah
01:36 PM exp: at the moment I also have been told that it sings at certain power levels
01:37 PM cehteh: switching PS?
01:37 PM exp: so I'll be doing a fair amount of experimenting with adding some higher ESR caps in parallel to add a little damping to the hot side
01:37 PM exp: yes but that's relatively well rejected, it's a load with high transient requirements
01:37 PM exp: so I suspect there's some harmonic current developing
01:37 PM exp: smoothes out over 20% power apparently
07:02 PM rue_ is now known as rue_mohr
08:48 PM specing_ is now known as specing