azonenberg changed the topic of #scopehal to: libscopehal, libscopeprotocols, and glscopeclient development and testing |,, | Logs:
juli969 has quit [Quit: Nettalk6 -]
<azonenberg> monochroma: also MEAD boards being assembled tonight. Jes is on the mill right now milling out the MMCX slots so the connectors actually fit
<azonenberg> a 1.45mm carbide endmill works great on fr408, and it looks like we only have to take off 50 μm from each side of the slot to make it fit
<monochroma> azonenberg: oh yeah, did you have any photos of what the issue there was?
<azonenberg> the issue is pretty simple, the connector slips into a U-shaped slot on the edge of the board
<azonenberg> the slot is about 75 μm skinnier than it should be
<azonenberg> so the connector won't fot
<azonenberg> fit*
<azonenberg> i think it's cutter compensation mismatch on my end vs oshpark's
<monochroma> :o
<azonenberg> i have the edge cut line right at the copper edge
<azonenberg> the actual board has about 50 μm of soldermask between the edge of the copper and the actual pcb edge
<azonenberg> i was actually gonna contact them about it once i had some closeup pics
<lain> oh man lol, I hope they drill those GND holes before they mill the slot
<azonenberg> lain: yeah those are vias so everything is drilled first, then they mill edges at the very end
<azonenberg> you can actually see bits of edge plating hanging free
<azonenberg> since the copper was just kinda ripped out
<azonenberg> which is fine, that's not the only ground connection for the pad
<azonenberg> i didnt actually even need it plated but oshpark plates all holes that go through copper
<azonenberg> there's no way to do a NPTH through copper
<azonenberg> Sooo there's no sign of the LSHMs
<azonenberg> guess i cant finish the MEAD boards today
<azonenberg> They're in my digikey order history with the rest of the parts, no record of being backordered
<azonenberg> but they're not in my connectors bin or the MEAD project bin
<azonenberg> So either digikey screwed up or i threw them out with the packaging when i opened the box
<azonenberg> I check pretty carefully so i think the latter is unlikely
<monochroma> you don't verify orders with what was received?
<azonenberg> I should
<azonenberg> normally all i do is open up the box then unroll all of the shredded paper to make sure nothing is left in it
<azonenberg> next time i wont make that mistake :p
<monochroma> i usually do a quick check of the included parts sheet against all the parts
<azonenberg> right side looks pretty good, left side has significant extra material
<azonenberg> so looks like a) registration between routing and copper is off by ~150 um and b) the hole is smaller than it should be
<azonenberg> i measure 3369 μm off the photo for total cutout width, the gerbers have 3450
Degi has quit [Ping timeout: 264 seconds]
Degi_ has joined #scopehal
Degi_ is now known as Degi
electronic_eel has quit [Ping timeout: 246 seconds]
electronic_eel has joined #scopehal
<azonenberg> Sooo
<azonenberg> the Murata DC-DC module i put on the MEAD+INTEGRALSTICK board - same one i'm using like four of on MAXWELL - is bigger than it looks
<azonenberg> it baaaarely fits on that board
<azonenberg> almost knocks a resistor off
<azonenberg> Aaaand that board is reworked and ready to go. Put the BLONDEL testbed away for a little bit to free up bench space
<azonenberg> Top side of the MMCX-to-MEAD board is assembled, bottom side is pending arrival of the LSHM connector
<azonenberg> MEAD board itself is also pending arrival of the LSHM
<azonenberg> Which should be arriving around the same time as the sfp-to-sma
<azonenberg> The MEAD+INTEGRALSTICK board looks good. after some fighting to align the tiny WLCSP power switch (not going to be using that on MAXWELL) i got it in place
<azonenberg> and the SFF connector fits properly and everything looks good there
<azonenberg> so I think that's about all i can do on the assembly/test side until the LSHMs or the SFP-to-SMA boards come in
<azonenberg> Still need to do lots of MAXWELL design review
<azonenberg> oh and then the v0.9 probe is in but i didnt have time to assemble today
monochroma has quit [Ping timeout: 246 seconds]
monochroma has joined #scopehal
katharina has joined #scopehal
<katharina> azonenberg: Will we need enums as preference values in the near future? E.g. something like how much metadata to emit to the session file etc
katharina has quit [Quit: Lost terminal]
juli969 has joined #scopehal
juli969 has quit [Quit: Nettalk6 -]
<_whitenotifier-f> [scopehal-apps] azonenberg reviewed pull request #122 commit -
<_whitenotifier-f> [scopehal-apps] azonenberg reviewed pull request #122 commit -
<_whitenotifier-f> [scopehal-apps] azonenberg reviewed pull request #122 commit -
<_whitenotifier-f> [scopehal-apps] azonenberg reviewed pull request #122 commit -
<_whitenotifier-f> [scopehal-apps] azonenberg commented on pull request #122: Basic implementation of the preferences dialog -
<_whitenotifier-f> [scopehal-cmake] azonenberg commented on pull request #9: Optional documentation/applications/examples -
<_whitenotifier-f> [scopehal-cmake] azonenberg closed pull request #9: Optional documentation/applications/examples -
katharina has joined #scopehal
<_whitenotifier-f> [starshipraider] azonenberg pushed 1 commit to master [+0/-0/±1]
<_whitenotifier-f> [starshipraider] azonenberg 0ff5e27 - Added missing AC coupling cap on OCXO output
<_whitenotifier-f> [starshipraider] azonenberg pushed 1 commit to master [+0/-0/±18]
<_whitenotifier-f> [starshipraider] azonenberg e45bbdf - Changed OCXO output termination to series instead of parallel
<_whitenotifier-f> [starshipraider] azonenberg pushed 1 commit to master [+0/-0/±1]
<_whitenotifier-f> [starshipraider] azonenberg f52f440 - Continued design review. Verified all thermal pads.
<azonenberg> monochroma: latest ETA on sma-sfp boards is saturday FYI
<azonenberg> finally got a tracking update
katharina has quit [Ping timeout: 264 seconds]
bvernoux has joined #scopehal
<bvernoux> hi
<bvernoux> For those interested I have done an other design of TRL Board v0.3 ;)
<bvernoux> I think this one will be the good one ;)
<bvernoux> using 0402 parts too
<bvernoux> OSHPark will sponsorize me also to produce it with OSHPark 4 Layers
<bvernoux> as such board is quite expensive with 4 Layers @OSHPark > 100USD for 3 boards
<bvernoux> There is special silkscreen around pin to easily solder it with other connector ;)
<bvernoux> That shall not impact performance a lot as it is very thin
<bvernoux> I can always remove this silkscreen too in worst case ;)
<bvernoux> I will compile all my tests results with different SMA connectors to show how that impact RF
<bvernoux> it is clear that cheap SMA connector have huge impact ;)
<bvernoux> so far the best is BELFUSE 142-0771-831 I need to figure out what is the issue with SouthWest Microwave connectors as they shall be even better in theory
<azonenberg> Have you tried the amphenol one i like?
<bvernoux> no
<azonenberg> They look almost identical
<bvernoux> It seems very similar to BELFUSE 142-0771-831
<bvernoux> will be interesting to compare them ;)
<bvernoux> anyway BELFUSE 142-0771-831 price is clearly winner so far
<bvernoux> when bought directly from Powell Inc
<bvernoux> 410USD (without VAT / Shipping) for 100 connector is nice ;)
<bvernoux> clearly a very good deal except they are pretty hard to solder so far
<bvernoux> maybe I will try the variant with legs (PCB hole) to easily align it
<bvernoux> but it is much more expensive ...
<bvernoux> azonenberg, the one you are using is Amphenol RF
<bvernoux> 901-10511-3 ?
<azonenberg> yes that sounds right
<bvernoux> ok
<bvernoux> azonenberg, do you have news on prob stuff ?
<bvernoux> or other stuff ?
<azonenberg> status update is as follows
<azonenberg> v0.9 probe pcb came in a day or so ago, assembling it is on my agenda for after work today
<azonenberg> integralstick-to-MEAD single pod host board arrived, is assembled, and works in that it supplies 5V power to the integralstick
<azonenberg> not tested further because i don't have a MEAD pod to plug into the other end of the cable
<azonenberg> MEAD MMCX input board is assembled top side only
<azonenberg> bottom side is on hold pending arrival of the LSHM series connectors which should be coming... tomorrow i think?
<azonenberg> main MEAD board is also here and on hold pending arrival of the LSHM
<azonenberg> enclosure prototype is here, fit verified against the boards
<azonenberg> MAXWELL schematic is done, review in progress. Probably will take the rest of the week before i can start layout
<azonenberg> SFP+ to SMA probe fixture is in transit, boards expected to arrive saturday
<bvernoux> ha nice lot of stuff
<azonenberg> bvernoux: did you see the MAXWELL ratsnest/component render
<azonenberg> ?
<bvernoux> it is HUGE ;)
<bvernoux> It will be a real headach to validate all
<bvernoux> and requires specific board to test each functions ...
<azonenberg> yes that. The board will be almost exactly the area of a microATX PC motherboard
<azonenberg> but slightly rectangular rather than square
<azonenberg> approximately 8x11 inches instead of 9.6 x 9.6
<bvernoux> for ultra complex boards with tons of RF stuff ;)
<bvernoux> I have seen interesting things about using internal PCB layers
<bvernoux> especially to keep the board compact
<bvernoux> with good isolation
<bvernoux> it is used on SM200B/C see video here
<bvernoux> It is interesting
<azonenberg> well the nice thing is this is almost all digital
<bvernoux> they have done some filters in different layers with special separation
<azonenberg> since its just a logic analyzer
<bvernoux> yes but for very fast digital stuff such hint are also very useful to gain space ;)
<azonenberg> Tentative stackup is L1/L3 signal ref L2 ground, L6/L8 signal ref L7 ground
<azonenberg> then L4/L5 power
<bvernoux> especially if you need to add some filters
<azonenberg> with L4-L5 dielectric FR408HR and the rest RO4350B
<bvernoux> ha interesting
<bvernoux> mixing different substrate
<azonenberg> well rogers is expensive and i dont need a low loss material between power planes
<bvernoux> yes clearly a big waste of money for power planes ;)
<azonenberg> so it seems like an easy way to cut the cost by 5-10% and not materially impact performance
<bvernoux> especially on such big PCB the cost shall be already very high
<bvernoux> and it will probably needs some revisions until it is "perfect"
<azonenberg> why do you think i've spent like four days on design review already and havent even started layout?
<azonenberg> the hope is to get it right the first time
<bvernoux> hehe yes to avoid as much as possible to do lot of PCB revisions ;)
<azonenberg> the bare board will cost me about $1900 for five pieces based on approximate quotes from multech based just on design rules
<bvernoux> what the maximum frequency expected for MAXWELL ?
<azonenberg> my digikey cart for the parts is about $1400
<azonenberg> for one system not counting any of the probes
<azonenberg> maximum saple rate? or what
<bvernoux> yes expensive ...
<azonenberg> 92 channels at 1.25 Gsps, 4 at 10 Gsps
<bvernoux> ha ok
<azonenberg> the RAM is DDR3 1600, then there's 40G ethernet to the outside world on the back
<azonenberg> with a higher speed grade fpga it may be possible to push the ram to 1866
<azonenberg> i'm only putting on a -2
<bvernoux> DDR3 is tricky IIRC more than DDR
<bvernoux> DDR4
<azonenberg> well the good news is i'm using a full SODIMM not loose chips. so a lot of the termination is done for me on the module
<bvernoux> But i'm confident you will produce something really amazing I hope it will be good on the 1st version
<bvernoux> it will be tricky to debug 10GSPS and even more for the 40GSPS stuff
<bvernoux> debug=validate ...
<azonenberg> well the 40G is a QSFP so four 10G lanes
<bvernoux> ha ok
<azonenberg> and they'll be fairly short differential runs straight from the fpga to a qsfp socket
<bvernoux> A good spectrum analyzer can help a lot to check RF reflection ;)
<bvernoux> or it will requires 100GSPS scope ;)
<bvernoux> anyway good test is to check the speed and stability of the links
<azonenberg> yes. also i can put a loopback QSFP in and do eye measurements on the fpga
<bvernoux> ha yes
<azonenberg> xilinx has an internal ber tester that will be helpfujl
<bvernoux> are you sure it will cope with 10GSPS to do eye diagram ?
<bvernoux> ha great
<azonenberg> its part of the serdes ip
<azonenberg> it only does a ber eye
<bvernoux> ha nice
<azonenberg> so you vcant see the outside of the eye, only the opening
<bvernoux> yes but it is enough
<bvernoux> Do you have SFP+ switch ?
<azonenberg> I have a nexus 3064x :p
<azonenberg> $500ish on ebay, amazing deal
<azonenberg> that thing probably cost $20K or more new
<bvernoux> ha yes great
<azonenberg> google says $28K
<bvernoux> with tons of ports ;)
<azonenberg> Yeah. But mostly 10G
<azonenberg> only four 40G
<azonenberg> so i will run out very quickly if i start using 40G stuff
<azonenberg> initial firmware will use a fanout cable and only run the link at 10G
<bvernoux> azonenberg, maybe you will have an offer from Lecroy/Keysight to buy your buisness ;)
<azonenberg> with the other three channels unused
<azonenberg> bvernoux: obviously i'd say no
<bvernoux> what they do not have is nice open GUI
<bvernoux> about technology they are very far
<miek> do they want a nice open GUI?
<azonenberg> We do :p
<miek> i don't know much about lecroy, but keysight/NI/etc. seem to like their closed systems :p
<azonenberg> yeah i think the image of a tek, a keysight, and a lecroy on one lab bench all interoperating is gonna be pretty scary to them
<miek> haha, yeah :)
<azonenberg> In one gui
<bvernoux> yes ;)
<bvernoux> their main issue is they are so closed
<bvernoux> and so expensive too ;)
<bvernoux> everyone speak about RF everywhere with 5G and other stuff but it is clearly not for little company ;)
<bvernoux> great
<bvernoux> OSHPark just gave me a coupon to pay the TRL Board v0.3 (PCB 1.6mm) 4 Layers OSHPark
<bvernoux> They are amazing to support open source stuff and RF stuff too
<bvernoux> It is 2nd time they "sponsor" PCB for RF tests
<monochroma> :D
<bvernoux> azonenberg, A thing I would like to test is ISIG ;)
<bvernoux> it is must have
<bvernoux> Immersion Silver then Immersion Gold
<bvernoux> without bad effect of ENIG or ENEPIG
<bvernoux> skin effect on RF microwave
<bvernoux> so RF loss is crazy low
<bvernoux> not far from pure copper
<bvernoux> without drawback of Immersion Silver or Pure Copper with oxidation...
maartenBE has quit [Ping timeout: 264 seconds]
maartenBE has joined #scopehal
<Degi> Oh neat
<Degi> Tbh we could just do immersion gold on immersion silver pcbs heh
<Degi> Immersion YBCO
<bvernoux> Interesting there is a live chat on Discord about launch of vapor phase one on Crowd Supply
<Degi> Vapor phase what?
<bvernoux> It will be fully open source/open hardware
<Degi> Soldering furncace?
<bvernoux> yes to solder with perfection ;)
<Degi> Ah nice
<Degi> Hm yes good for even heating
<bvernoux> it use very expensive Galden
<Degi> Not that expensive...
<Degi> Hmh well 189 € for 500 ml... But there could be cheaper
<bvernoux> it seems hard to find cheaper
<bvernoux> as it seems only 2 company in world are producing it so they do what they want with price ;)
<Degi> What, alibaba has a MOQ of 20 kiloamperes?
<Degi> I remember seeing it cheaper somewhere... might misremember
<bvernoux> HT200 is not good IIRC for soldering
<Degi> Alibaba isnt much cheaper either
<bvernoux> the usable ones are => Galden LS 230, XS 235, and HS 240
<Degi> Oh
<bvernoux> 200 is too low
<Degi> Well for SnBi
<Degi> Maybe somebody could make an open-source chemical factory
<bvernoux> especially for Lead Free which needs high temperature
<miek> i'm not sure i want to know the chemsitry involved in making that stuff, let alone be near it :p
<miek> have you seen the msds!?
<Degi> You probably just mixed some fluorine gas with organic stuff
<Degi> Link?
<bvernoux> glurps 2899USD Vapor Phase One
<bvernoux> I will wait to have bigger buisness in PCB/Electronic stuff before to buy it ;)
<Degi> 20 $ raised huh
<bvernoux> I prefer to use other factory to build & solder my PCB ;)
<bvernoux> So far it is a good deal except it take time ...
<Degi> What is mg Hg in terms of vapor pressure lol
<Degi> Nice msds
<Degi> "Solvay Acceptable Exposure Limit 1000ppm" Sure, who tested that
<miek> overheating it would really ruin your day :p
<Degi> Hmh depends
<Degi> Throw some baking soda and water at it
<bvernoux> miek, do you plan to buy one ?
<Degi> Also what does "Use per 100 prints" mean? Where does it go?
<Degi> Not sure if youd want to release that to atmosphere
<miek> no, no immediate plans.
<Degi> Thats as irresponsible as those "canned air" organofluoride things
<miek> the t-962a does fine for my needs right now
<Degi> (Well once I've sprayed that into a mass spec and may have inhaled the exhaust, but its not much)
<bvernoux> interesting but galden is patented by solvay ...
<bvernoux> it is why it is so expensive
<bvernoux> ...
<Degi> I wonder how it works
<Degi> Cause apparently its pressure independent?
<bvernoux> yes
<bvernoux> the magic is the temprature is ultra stable
<bvernoux> and it is like a cloud
<Degi> How does that work
<miek> i dunno, it's not that bad now that people are selling smaller amounts
<bvernoux> the liquid is hot and then it go in vapor at the expected temperature
<miek> i spent roughly the same on a pot of good solder paste...
<bvernoux> HS 240 => 240°C
<bvernoux> IIRC
<Degi> But how exactly does that happen, does it have a very sharp vapor pressure curve
<bvernoux> then you refresh and it became like water
<bvernoux> They use it for aerospace stuff
<bvernoux> to be sure soldering is perfect
<bvernoux> and nothing is overheating
<Degi> But like how is it pressure independent
<bvernoux> as it cannot overheat
<bvernoux> it is heavier than air
<bvernoux> so it keep at a constant height
<bvernoux> until you overheat IIRC
<Degi> So at the temperature the vapor pressure goes from almost zero to very high?
<bvernoux> it is heavy "gaz" in fact
<bvernoux> I have not read anything about high pressure
<azonenberg> Degi: the big thing is that you're heating the liquid
<Degi> gaz?
<bvernoux> yes ;)
<Degi> Hmh yes the liquid vaporizes, condenses
<Degi> But how is that different from like boiling water?
<azonenberg> and as long as there's still liquid in the tank, the gas can't be hotter than the boiling point of the liquid
<Degi> Somebody told me that that'd work at lower pressures too, like with a vacuum pump attached
<azonenberg> it stays right at the boiling point, condensing on anything it can
<azonenberg> including the board
<bvernoux> Degi, it is the same but at 240°C ;)
<sorear> is the pressure assumed to be atmospheric?
<bvernoux> so perfect for soldering
<Degi> Like what if your solder chamber is at 10 bars? Does it get a lot hotter then?
<bvernoux> Degi, it is intended to work at normal pressure IIRC
<Degi> Otherwise I doubt that it can be run with a vacuum pump, since that usually drastically reduces the boiling point
<bvernoux> at sea level ;)
<bvernoux> homogeneous heat transfer which never exceed 240°C
<Degi> Like cant you just use Tridecane
<bvernoux> which is not the case with hot air, soldering iron or hot plate ....
<Degi> Like tridecane is 2-4 $ per kg on alibaba
<bvernoux> if you overheat galden it vaporize IIRC and you loose it ;)
<Degi> 232-236 °C boiling point
<Degi> Downside is its combustible but just dont light it on fire...
<bvernoux> yes boiling point is the number in the reference
<bvernoux> Galden HS 240 => 240°C
<Degi> Actually n-Tridecane might be worth a shot
<bvernoux> it is why Galden 200 is not interesting for soldering
<bvernoux> But each time you solder a board you loose 3gram IIRC
<bvernoux> if you compute 100Euros / KG
<Degi> Hmh maybe not if you let it cool down
<bvernoux> Degi, you always loose a bit
<bvernoux> it is wrote in their spec
<Degi> Where to
<bvernoux> I checked this one before
<bvernoux> As long as the bin is never opened during the soldering process, the consumption will not be more than about 0.5L per 100 prints
<Degi> I mean you could even use water
<bvernoux> So I suspect it is the same with Vapor One
<bvernoux> it is wrote
<bvernoux> Volume of Galden used per cycle: 0.49 ml (about $0.16)
<bvernoux> So not so expensive but the HW is expensive anyway
<Degi> I mean you could just compress water to 4 MPa and do soldering haha
<Degi> Or ethylene glycol at 0.3 MPa
<bvernoux> I doubt water is good for soldering even at high pressure ;)
<Degi> I mean water has very good heat transfer
<Degi> Like 2 kJ/g
<Degi> (for 100 °C)
<bvernoux> but for corrosion ...
<Degi> hmh yes thats a downside
<bvernoux> Galden has not that effect
<bvernoux> so it is not simple water boiling at 240°C ;)
<Degi> O.o
<Degi> What if you use glycerol at 200 mBar
<Degi> Thats very doable
<Degi> And its friggin cheap
<Degi> Hmm it might dissolve certain plastics
<Degi> Does Galden need to be washed off at the end?
<Degi> Though that would need active heater control
<Degi> Basically 1) vacuum out chamber, 2) heat glycerol until 200 mBar reached 3) cool down and wash
<sorear> i suspect glycerol is also pretty corrosive with the reactive -OH
<sorear> OTOH aliphatic alcohols are significantly less acidic than water
<Degi> Hmh#
<sorear> glycerol will definitely corrode, say, sodium at room temperature, no idea if it attacks copper at 240C
<Degi> Huh
<Degi> I mean probably not very much? idk
<sorear> on the other other hand, chelate effect
<Degi> Does that also cause corrosion
<sorear> probably, since it favors the ionic phase
<Degi> Hmh copper?
<sorear> chloride or cyanide would complex pretty strongly
<Degi> I mean even pure HCl doesnt etch copper
<Degi> Well it needs an oxidant like oxygen from air
<bvernoux> Degi, Galden does not need to be washed or anything at end
<bvernoux> Degi, it is the magic ;)
<Degi> Hmh where does it go
<bvernoux> it go back like awter
<bvernoux> water
<bvernoux> when refreshing
<bvernoux> there is a video about it
<bvernoux> what is lost is some micro wavelet of galden IIRC
<bvernoux> also it seems safe for human if you eat it ;)
<Degi> Biphenyl has a good boiling point
<electronic_eel> azonenberg: do the UARTs to the pods on MAXWELL have EDS protection?
<electronic_eel> I didn't see any, but I still have a strong feeling of being overwhelmed by the complexity...
<azonenberg> electronic_eel: they do not, good catch
<azonenberg> ESD protection is quite a few lines down on the review checklist so i havent got there yet
<electronic_eel> also I'd add some 33 ohms series resistors, just for good measure. maybe use some resistor packs for it
<electronic_eel> the *_present signals also don't have ESD protection
<azonenberg> Noted
<electronic_eel> what are the 300 ohms resistors between the pod power and the pods for?
<azonenberg> those aren't resistors
<azonenberg> those are 300 ohm @ 100 MHz ferrite chis
<azonenberg> chips*
<electronic_eel> ah! yeah, FBnn
<azonenberg> To minimize radiated emissions on the pod cables
<azonenberg> I'm trying to actually put some effort into EMC on this thing
<electronic_eel> 300 ohms wouldn't have worked well with the pods...
<electronic_eel> I'm confused by all the different I2C busses and them constantly changing their name from sheet to sheet
<electronic_eel> you need that many, no doubt about that
<electronic_eel> but wouldn't it help to keep their names constant across the sheets, so you can follow them better and better tell them apart?
<electronic_eel> that will also give them consistent names in the layout file
<azonenberg> i did this via a bottom-up design and some net names were assigned in sub-sheets then, for example, two different sheets' buses got put on the same physical bus
<azonenberg> and renaming dozens of net connections in the sub sheets seemed error prone
<azonenberg> the final net names are hierarchically scoped
<electronic_eel> can't you do a simple rename, just with search & replace? I don't consider that error prone
<electronic_eel> giving the same thing 3 different names on different sheets seems more error prone to me
<electronic_eel> where are the pullups for the probe and power / ram i2c busses?
<_whitenotifier-f> [starshipraider] azonenberg pushed 1 commit to master [+0/-0/±6]
<_whitenotifier-f> [starshipraider] azonenberg e9def1e - Continued schematic review
<azonenberg> electronic_eel: very possible they're missing, checking for missing pullups is a few steps down on the review
<azonenberg> i'll add a note to look
<electronic_eel> searching for stuff like that would be easier if the bus names were consistent
<azonenberg> yeah let me see what i can do on that
<_whitenotifier-f> [starshipraider] azonenberg pushed 1 commit to master [+0/-0/±10]
<azonenberg> Updated, new PDF uploaded
<_whitenotifier-f> [starshipraider] azonenberg 70188ab - Renamed I2C buses to be consistent
<electronic_eel> much better, thanks
<electronic_eel> my kicad always takes about 5 seconds to load the schematics
<electronic_eel> never had a schematics this complex before...
<electronic_eel> the supercap for timekeeping is something that will age. but seiko doesn't want to tell in what kind of timeframe (ds: "long life span, high reliability")
<electronic_eel> it is a quite small smd package, I don't know if the user can be expected to change it in 10 years (or when the thing will fail)
<Degi> Maybe mark it on the PCB haha
<Degi> "Place a 3.3 V supercapacitor here of some dozen mF if timekeeping fails or fluids leak out"
<electronic_eel> on the other hand it is nice that it is so mall
<electronic_eel> I don't know how it will fail, just no capacity left, or will it have a huge leakage current?
<Degi> Maybe itll have leakage fluid
<electronic_eel> if it is just the capacity that goes down, we could add a pin header and the user could plug in another cap
<sorear> why does it have a RTC?
<Degi> Good question
<electronic_eel> to tag the samples with a rough time
<Degi> Cant it do ntp
<electronic_eel> for more precise timing ntp, ptp and gps is planned
<Degi> Or will the time need to be entered at the front panel
<electronic_eel> that wouldn't be nice ux. you'd probably set it from the host instead
<electronic_eel> there is a safety datasheet on digikey. "solute: Tetra alkyl ammonium salt, solvent: hetero-oxide"
<electronic_eel> don't know what that is, but it sounds a bit to me like it could leak
<Degi> We could use lithium thionyl chloride batteries
<electronic_eel> given enough time
<Degi> The oxide could be a solid maybe
<electronic_eel> I had good experience with the regular CR2032, they are cheap and easy to replace and work well for >5 years with a simple RTC
<electronic_eel> only downside is that they take a lot of space
<Degi> yes
<electronic_eel> but I've seen some mainboard manufacturers plugging them in with a small jst connector and sticking them someplace on the pcb with double sided tape
<electronic_eel> the battery is shrink wrapped and the cable welded on
<Degi> ah these
<electronic_eel> I guess you can buy them that way, so you just need the connector on the pcb
<electronic_eel> yes, exactly
<Degi> 19 amp h
<azonenberg> yeah i dont plan to use the bbram for settings or other critical stuff
<Degi> Huh 2.5 MJ/kg
<azonenberg> the battery is just to persist time across short shutdown cycles
<azonenberg> without hammering the ntp server
<electronic_eel> but if it begins to leak after a few years...
<azonenberg> we might even end up DNPing on the final build and just having the diode+cap to supply vbat when it's on
<electronic_eel> I'm not very convinced by the ""long life span, high reliability" language in their ds, no actual numbers anywhere, that sounds fishy to me
<Degi> Shorting out thionyl chloride ones sounds fun
<electronic_eel> how about replacing it with a molex kk connector for plugging in a lithium battery like the ones degi linked?
<azonenberg> i think it makes more sense to keep it as designed, but with the assumption of DNPing down the road?
<azonenberg> i really dont even think it will be used at all
<Degi> Check out the # available, this sounds like fireworks
<electronic_eel> azonenberg: hmm, why do you have it in there, then?
<azonenberg> Because a DNP is less work than bodgewires :p
<azonenberg> same reason i hooked up the RMII between the FPGA and STM32
<azonenberg> I have no immediate plans to use it, but it might be needed in the future
<electronic_eel> yeah, but it might come in handy to have for some feature or workaround that we do not know about now
<electronic_eel> but I think having a battery then, which can be user replaced, is better than having a leaking supercap
<Degi> Why does digikey have several hundred kiloampere-hours of LiSOCl2 batteries on storage
<azonenberg> electronic_eel: batteries can leak too, and this thing is easy to rework if needed
<azonenberg> its also a hermetically sealed ceramic+metal package like a crystal oscillator
<electronic_eel> Degi: the person responsible for fire hazard precautions at Digikey wants something to do
<azonenberg> i doubt it would leak
<azonenberg> fail, perhaps, but not leak
<Degi> Thats like a few dozen kg of lithium haha
<electronic_eel> Degi: extinguish the fire with water
<Degi> big boom
<electronic_eel> azonenberg: I'd prefer the pin header solution, I think it is more flexible for something that you don't know the exact usecase for
<azonenberg> electronic_eel: ok fine... does it have to be a KK or can i just put down two 100 mil spaced pins and we'll go from there?
<Degi> That should be fine if you mark polarity on silkscreen
<azonenberg> Ok
<azonenberg> So then the user can populate the cap+diode, battery, or any combination thereof
<electronic_eel> just the 2.54mm pin headers, you can solder in kk's or anything similar later on
<azonenberg> will also double as a test point for the cap
<_whitenotifier-f> [starshipraider] azonenberg pushed 2 commits to master [+1/-0/±2]
<_whitenotifier-f> [starshipraider] azonenberg 102212d - Added 2 pin header to VBAT rail for optional external battery
<_whitenotifier-f> [starshipraider] azonenberg 115c913 - Added autosave to gitignore
<electronic_eel> ok, enough review for me tonight, will go to bed now
<Degi> Huh, CNC devices for 200 €
<electronic_eel> hrmpf, can't shut off my brain from review mode this fast, I keep thinking about the schematics and still find stuff, even with no pc on
<Degi> Heh yees
<electronic_eel> I suggest to add polyfuses to the fan headers. a fan could be blocked or otherwise damaged and create a short on the 12v0
<azonenberg> electronic_eel: They have PWM and tachos
<azonenberg> if the tach shows no motion i'll stop sending PWM
<azonenberg> i guess a polyfuse wouldnt hurt on top of that
<electronic_eel> yes, but that won't help if the fan bldc driver is burnt
<azonenberg> that *is* a pretty high current rail
<electronic_eel> I had that on a pc at a customers site, the wire to the fan burnt the isolation
<azonenberg> 500 mA sound like a good polyfuse rating?
<electronic_eel> that is fine, the fan should never draw that much
<azonenberg> Exactly my thought, the one in my cart is rated for 170
<azonenberg> so thats a comfortable margin
<electronic_eel> ok, now I try again to get some sleep
<azonenberg> 507-1799-1-ND?
<azonenberg> 500 mA trip, 250 mA hold
<electronic_eel> yeah, that should work fine. I use some bel fuse polyfuses too and they worked well
bvernoux has quit [Quit: Leaving]
<_whitenotifier-f> [starshipraider] azonenberg pushed 1 commit to master [+0/-0/±4]
<_whitenotifier-f> [starshipraider] azonenberg 1e78cf7 - Added polyfuses on 12V0 rails for fans
<Degi> Huh, fans dont have current limiting?
<azonenberg> Degi: dumber fans as opposed to BLDC ones are literally just a dc brushed motor hooked up to the 12v rail