<miek>
mine's the Agilent 1156A. as far as i can tell it's not a PMK rebadge, but i could be wrong. it does look like they might've got all the accessories from PMK though
<azonenberg>
yeah the pico ta061 looks like an in house design with at least mostly PMK accessories
<azonenberg>
now that i have four of them and have used them a bunch, i'm... actually less than impressed
<Degi>
How linear does time to digital need to be?
<Degi>
Can it be somehow calibrated?
<azonenberg>
Degi: as long as we can convert code values back to times, linearity doesn't matter. Stability and reproducibility does
<azonenberg>
the internal transfer function is of no consequence
<azonenberg>
could just be an arbitrary polynomial/piecewise linear or even table based conversion
<Degi>
Hm i
<Degi>
okay
Degi_ has joined #scopehal
Degi has quit [Ping timeout: 240 seconds]
Degi_ is now known as Degi
<azonenberg>
electronic_eel: Thoughts on the OKL-T/3-W12P-C for the +7V rail?
<azonenberg>
it's moderately high current since it supplies the gain stage, mcu, etc. More to the point, it has plenty of capacity to run 4 channels (I'm trying to oversize the PSU for the 4-channel anticipated load)
<azonenberg>
i like the prebuilt LGA modules for convenience, and it's only $6
<azonenberg>
3A capacity is overkill but who cares
<azonenberg>
oh wait
<azonenberg>
it maxes out at 5.5V output
<azonenberg>
there goes that idea
<azonenberg>
it's still a nice psu
<Degi>
Why does the +7 V supply the MCU
<azonenberg>
there's going to be a 3.3V LDO off that
<azonenberg>
i don't see the point in having a dc-dc just to run the mcu
<Degi>
Hm I guess the MCU doesn't need much current, right?
<azonenberg>
and LDOing off the 7V rail is more efficient than the 12
<Degi>
Oh neat, that's a complete buck converter module
<azonenberg>
yeah but it can't do +7. maxes out at 5.5 output
<azonenberg>
The mcu is just for the test board anyway, the full system will be all FPGA, so i don't want to waste a lot of engineering on the testbed
<Degi>
Should be able to do at least 9 V at 12 V in
<azonenberg>
I'm eyeing the TPS54227 right now
_whitelogger has joined #scopehal
<azonenberg>
decided to use a lt3042 for the mcu rail just to keep the design internally consistent. Total overkill but it's not worth saving the $4 for a board i'm only going to ever make one or two of
tnt has joined #scopehal
<electronic_eel>
hehe, the lt3042 for the mcu is really decadent
<azonenberg>
does the supply design look good in general so far?
<electronic_eel>
but I don't see any issues with the psu section at the moment
<azonenberg>
i mean obviously not done
<electronic_eel>
I don't have experience with the TPS54227, but the TI switchers are usually nice to work with
<electronic_eel>
did you use their we bench thing to select it for low noise?
<azonenberg>
No
<azonenberg>
i figured noise didnt matter a ton as it was an intermediate rail
<azonenberg>
basically my selection process was, digikey search buck converters with input Vmax >= 12V and output Vmax >=7V
<azonenberg>
exclude PTH, exclude brands i've never heard of
<electronic_eel>
the ldos are mostly effective for lower freq noise, the higher freq stuff usually has still to be filtered if it matters
<azonenberg>
it was the first one that looked decent
<azonenberg>
oh and i added a 500 mA minimum output current filter too
<azonenberg>
if you look, i have a ferrite between the buck and the ldo
<electronic_eel>
but it doesn't switch to pfm mode to save energy in low current situations, so it is well suited for filtering
<azonenberg>
which forms a C-L-C pi low pass filter with the input and output caps
<azonenberg>
this was suggested in the lt3042 datasheet as a means of improving PSRR when postregulating a smps
<electronic_eel>
there is just one thing I don't like: you drawing components going to gnd upwards
<electronic_eel>
that is really confusing to me
<azonenberg>
lol, i try not to do it too much but sometimes in a dense schematic it's the only real way to make things fit without taking up a lot more space
<azonenberg>
if i have room once the psu is done that was already on my list of things to prettify
<azonenberg>
Prior to the design review i usually do a "make it look nice" pass on the schematic moving blocks around, adding text annotations to subsystems, etc
<azonenberg>
oh before i forget
<azonenberg>
i owe you pics of the zs1500 and accessories still, right?
<electronic_eel>
yeah, pics of those would be nice
<electronic_eel>
wouldn't it make sense to put a bulk capacitance on the 12v in, like a polymer cap 200µF or similar?
<azonenberg>
in the interests of time i wont include the color clips, copper tape, spare tip needles, etc
<azonenberg>
Yes there will probably be something there. Once i've done the whole psu and have a better estimate of peak ripple etc probably
<electronic_eel>
just put one in with "TBD" as value
<electronic_eel>
no pictures of color clips necessary ;)
<electronic_eel>
a bit bulky this probus thing, but ok
<azonenberg>
you mean how far it sticks out, or how big the connector itself is?
<azonenberg>
that body contains active electronics, it's not just a connector
<azonenberg>
what, i'm not sure as i havent opened it up
<azonenberg>
but i've seen teardowns of other probus probes
<electronic_eel>
the whole width of the connector case
<azonenberg>
as a minimum it's a regulator from +/- 12V intermediate rail from the scope down to whatever the probe buffer natively uses, plus a 24c discovery eeprom
<azonenberg>
its not unreasonable compared to other active probes i've seen
<azonenberg>
this is the tip of the zs1500 showing the ground and signal contacts. I am thinking of copying that relative placement in my probe in order to be fully compatible with tetris accessories, so this may require a bit of an enlargement of the probe pcb in the width axis
<azonenberg>
but as a minimum i need to pull the ground back a little bit in order to have the z-ground be coplanar with the trip
<electronic_eel>
yeah, I thought about if that would make sense to do on the passive probe to become compatible
<azonenberg>
i think that may actually slightly hurt signal integrity, but is a necessity for ergonomics to allow variable signal-ground spacing
<azonenberg>
the z-ground is excellent for that
<azonenberg>
i may also supply a second, closer ground socket. you can always use one or the other
<azonenberg>
here we have at top two accessories i basically never use. the topmost appears to be an adapter from a female pin socket to the machine pin socket on the probe
<azonenberg>
then the ground leaf
<electronic_eel>
now the interesting stuff
<azonenberg>
next is a rather odd one, it's an adapter from a 2 pin 100 mil header to the pin spacing of the tetris signal/ground
<azonenberg>
it's literally a square pin header wire bent to shape with a ground leaf cut out of brass soldered onto it, presumably to reduce inductance
<azonenberg>
never used it but it wasnt in a tube so i got a better pic
<azonenberg>
then the z-ground which i use all the time, and the ground blade which is my #2 most often used ground
<azonenberg>
at top are the two clips. they're e-z hooks, unsure if customized or vanilla but they look unmodified
<electronic_eel>
do they directly go into the probe? I thought the diameter of their pin is smaller
<azonenberg>
They go directly into the probe. they have square pin headers that seem to acceptably-ish mate the socket which is meant for a round machine pin
<azonenberg>
i didnt measure the female socket size but all of the round pins are 0.76mm diameter
<azonenberg>
the socket i am using on my probes will take i think 0.5 to 0.8
<azonenberg>
anyway, then there's a range of straight and right angle ground leads. these are something i think i could probably make myself cost effectively. literally a machine pin at one end, a length of flexible wire, and a female header
<azonenberg>
then the one at the bottom is a signal and ground heatshrinked together
<azonenberg>
with the right spacing and length ratio
<azonenberg>
i think they're literally hand made by a tech somewhere
<electronic_eel>
yeah, probably
<azonenberg>
PMK charges a whopping $28 for the 7cm flex ground lead i included in my beta accessory order
<azonenberg>
i'm sure i could make my own for a fraction of that, and i plan to in production. for beta units i'm using premade just to eliminate variables, plus to provide a basis for comparison when i make my own
<electronic_eel>
I think this is something a hobbyist can make for himself and a corp buys this stuff for $$$
<azonenberg>
the tips are $31 per 5 pack which is not unreasonable, i plan to only include one per probe since tips really don't wear out often and it keeps costs down
<azonenberg>
if somebody wants more they can buy more
<azonenberg>
the ground blade is $21 but it also includes custom cut sheet metal so for low volume i'd have a hard time beating that
<azonenberg>
the z-ground is $16 and is also bent to a specific shape and has a hardened... steel maybe? tip that i havent been able to source so i'll likely buy that
<electronic_eel>
what is the best way to probe two male headers 2.54mm pitch, signal and ground next to each other?
<azonenberg>
the flexible lead i think i can make cheaper
<azonenberg>
then they quoted me something crazy like $40 for the bipod positioner whcih i can get for $25 from tequipment
<azonenberg>
just with pico branding instead of pmk
<azonenberg>
standard square pin headers are, per google, 0.64 mm square with rounded corners, 0.8mm diagonal
<azonenberg>
so they will fit into the tip socket
<azonenberg>
you can then use one of the short flex grounds for the ground. I may make one exactly the right length for this purpose, since the shortest one pmk sells is a 5cm which is a bit overkill
<electronic_eel>
not only overkill, but also hurts si
<azonenberg>
Yes, exactly
<azonenberg>
i did not see a stock accessory for this purpose. the weird blade contraption is meant to fit into a *female* 2 pin 2.54 header
<azonenberg>
but nothing to mate with a male
<azonenberg>
Good news is that transmission line probes are far less sensitive to ground inductance than the old school passive probes
<azonenberg>
it does have an effect of course, but it's a lot less
<electronic_eel>
I often put male headers on boards, seldom female ones. so some female thing for the probe would be used more often
<electronic_eel>
but if you can directly plug the probe on top of a header it is a big improvement to classic scope probes
<azonenberg>
Yeah. For the beta units, you can use the... 7cm i think, pmk ground lead i ordered
<azonenberg>
and plug the signal contact directly in (after pulling out the tip needle, of course)
<electronic_eel>
I still want some kind of grabber. would you mind to ask your contact at the pmk distri for details about these "SMD Test Grabber, 0.8 mm Pin, black, Flex Nose"
<azonenberg>
honestly what i've found myself doing often with the zs1500 is just putting the probe on a bipod. leaning the needle against the signal conductor, and using either the z-ground to touch the ground or the flex ground to the pin
<azonenberg>
the only "details" they'll send me is the price
<azonenberg>
i can put one on my next pmk order, whenever that is, and see what it's like
<electronic_eel>
hmm, that is a weird kind of sales
<electronic_eel>
that they don't even have a photo or something
<azonenberg>
they're a distributor, like digikey but more specialized
<azonenberg>
they arent the designer of the part, they don't know any more than the manufacturer gives out
<azonenberg>
all they do is handle the importation and such, since pmk won't ship direct to me for some reason
<electronic_eel>
ok, so one should ask directly at pmk for it
<azonenberg>
Yes
<azonenberg>
but it's a $250 shipping charge so i probably won't be doing another pmk order until i am starting production of first round probes
<electronic_eel>
maybe I send them an inquiry directly
<electronic_eel>
they probably sell directly to german customers, so they will probably talk to me when I contact them from my work address
<azonenberg>
Very possible. I also don't speak enough german to able to have an easy time with their sales team
<azonenberg>
the lady who answered the phone spoke a decent bit of english, but there were still some hangups
<electronic_eel>
you speak german at all?
<azonenberg>
Ein bischen
<electronic_eel>
hehe
<azonenberg>
when i was at RU Bochum giving a talk to christoph paar's research group, everyone i ran into picked up on my terrible accent and switched to english after a couple of words
<azonenberg>
i did manage to order Orangensaft at the hotel restaurant entirely in German
<azonenberg>
but that was it ;p
<electronic_eel>
but that is a good start
<azonenberg>
i mean the students all spoke very good english, and my talk was in english
<electronic_eel>
most people I met from the us didn't speak a second language or at least tried at all
<azonenberg>
but i can't remotely hold a technical conversation in the language
<azonenberg>
I was homeschooled
<azonenberg>
My mom is a piano teacher who, before i was born, wanted to be an opera singer
<azonenberg>
so i had my choice of learning any language anyone had written a nontrivial opera in :p
<electronic_eel>
ah
<electronic_eel>
Wagner has some nice examples of nontrivial...
<azonenberg>
side note, the christmas market in Bochum was quite nice. Definitely a good time of year to be visiting Germany, although i probably should have packed warmer gloves
<azonenberg>
it was a quite nice experience overall - quite a small world in fact, turns out one of the students in his group has recently been to the US (i forget if academic or personal travel)
<electronic_eel>
yeah, some towns have nice christmas markets
<azonenberg>
not only the US, but the city a ten minute drive from me that i used to live in, and still (pre coronavirus) commute through on a daily basis
<azonenberg>
and had eaten at a crepe place downtown that i am quite the fan of
<azonenberg>
Lol
<electronic_eel>
that is quite a coincidence
<azonenberg>
what are the odds? i'm literally on the other side of the globe in a foreign country and the student i'm having dinner with has been to my former home town - a city with a population of 20000, not like seattle - and eaten at one of my favorite restaurants
<azonenberg>
we checked the dates and i had indeed been living there when she visited, although of course neither of us had any idea the other existed at the time
<azonenberg>
lol
<azonenberg>
re languages, americans generally don't need to because english is such a universal tongue
<azonenberg>
what i've found is there are two universal languages for tech: Mandarin and English
<azonenberg>
if you're in Asia, you speak Mandarin plus your native language
<azonenberg>
in Europe, you speak English plus your native language
<electronic_eel>
yes, Mandarin is the one I'm missing to speak
<azonenberg>
Some, of course, speak all three (if english or Mandarin is not the native)
<azonenberg>
I... am not one of them
<azonenberg>
"ni hao" is as far as i can go in spoken chinese
<azonenberg>
and i can read the ideographs for "big", "middle", and "exit"
<azonenberg>
if you ask me the stereotype of chinese people being smart is very true, you'd have to be some kind of genius to memorize all of those symbols :P
<electronic_eel>
in europe it is english + native for the better schooled ones. you still meet lot's of people that don't speak english, especially elderly
<azonenberg>
Yeah i'm talking in tech
<azonenberg>
locals who dont deal with the global economy are a different story
<electronic_eel>
in tech you have no chance to get anywhere if you don't speak english
<electronic_eel>
or at least read it well enough
<electronic_eel>
thanks for the probe photos btw
<electronic_eel>
oh, about the TPS54227 - it is only max 18vin. if you want to use the same part to create the negative voltage that is too low
<electronic_eel>
it might make sense to use one type of regulator for both to reduce BOM
<azonenberg>
too low? i dont understand
<azonenberg>
it's a single supply +12V input
<azonenberg>
i need an inverting regulator of some sort for the negative rail
<electronic_eel>
yes, but if you want to use the buck to create -7 too, the buck sees 19v vin
<azonenberg>
oh i was going to go direct from +12 to -7 if possible
<azonenberg>
but i havent looked into smps topologies for negative output in a while
<electronic_eel>
yes, of course. but the voltages add up then
<azonenberg>
i usually use charge pumps for negative rails but this design uses a fair bit more current than i've needed in the past
<electronic_eel>
you can use regular bucks to create negative voltages
<electronic_eel>
that is so common that there aren't many dedicated negative switchers out there anymore
<azonenberg>
oh?
<azonenberg>
how do you do it
<electronic_eel>
basically gnd of the switcher is the negative output
<electronic_eel>
and vout is gnd
<electronic_eel>
"basically vss of the switcher" is the better wording
<electronic_eel>
it is not a must to do it that way, but just an option that you might want to consider
<azonenberg>
will look into it
bvernoux has joined #scopehal
<Degi>
Oh didn't I link a chip for that earlier
<Degi>
I should've noted that down somewhere...
<Degi>
Tbh a resonant power converter (something like the popular ZVS circuit could be sufficient, maybe something with OTP, OCP) could be used with a transformer, it could make +- 7 V (assuming a fixed 12 V input) and have low power at higher frequencies than the fundamental, thus be less noisy and easy to filter with a resonant filter
<Degi>
For BGAs, do I want smaller or bigger pads?
<azonenberg>
bigger or smaller than what?
<Degi>
Hm 0.4 mm or 0.6 mm for 0.8 mm pitch BGAs
<azonenberg>
pitch is of no consequence when selecting pad diameter
<Degi>
Nvm 0.4 mm seems to be max I can dogbone
<azonenberg>
what matters is ball diameter
<Degi>
Oh
<azonenberg>
not all bgas of a given pitch have the same ball size
<awygle>
pitch might affect SMD vs NSMD pads
<awygle>
but within those categories ball size determines pad size
<azonenberg>
I have a table i usually use from TI that uses ipc recommendations from ball diameter -> NSMD land size
<Degi>
0.4 mm is ball dia
<azonenberg>
For 0.4mm balls you want 0.3mm land diameter
<awygle>
azonenberg: do you standardize on nsmd then? any cases where you'd prefer sdm?
<awygle>
*smd
<Degi>
Well 0.35-0.45 mm
<azonenberg>
yeah 0.4 nominal
<azonenberg>
awygle: i check vendor recommendations, but almost never use SMD pads
<Degi>
Whats the difference between nsmd and smd?
<azonenberg>
copper size, not mask opening size, is what matters when determining how much of a via you can fit etc
<awygle>
SMD is "solder mask defined" - copper is larger than soldermask opening
<awygle>
NSMD is "non solder mask defined" - copper is smaller than soldermask opening
<awygle>
you can fit more stuff between SMD pads because they're smaller copper pads, but they're perhaps more likely to rip off the board
<awygle>
or the tolerances may not be as good
<awygle>
depending on process etc
<azonenberg>
you mean between NSMD?
<awygle>
err, yes
<awygle>
sorry, i always fuck that up
<azonenberg>
and yes SMD pads have slightly better retention
<azonenberg>
but honestly if you have any concerns about that just glue the bga post solder
<awygle>
or pre solder, or during solder
<azonenberg>
I don't like pre/during solder glue because it inhibits self alignment
<azonenberg>
and thus any slight misalignment becomes permanent
<azonenberg>
if there was a glue that would cure during cooling post reflow, etc that would be great
<azonenberg>
but my thermal cure epoxy cures at ~150C so by the time solder melts it's well solidified
<Degi>
Hm though NSMD won't be really round then, right?
<azonenberg>
awygle: ok i take that back. i sometimes use "parasitic SMD"
<Degi>
Oh the ball diameter is 0.45 mm for the 256 package. In the file I found the PCB has 0.41 mm dia pads. Guess thats ok
<azonenberg>
if i have via-in-pad and minimum via + annular ring is larger than the bga land size
<azonenberg>
for 0.45mm balls you can get away with going down to 0.35mm lands
<Degi>
Good to know
<Degi>
Is there a downside to making the mask opening as big as the pad?
<azonenberg>
you dont want them equal because of potential misalignments
<azonenberg>
you can run into issues where the effective pad is shaped weird because the mask and pad dont line up
<Degi>
Hm wont NSMD pads be weirdly shaped too?
<azonenberg>
so you either want the copper bigger than the mask, or the mask bigger than the pad, by 50 um or so (depending on your fab's alignment tolerance, 50 is my default)
<Degi>
Because of the wire leading to the via
<azonenberg>
Yes. but that tends to not be a big issue
maartenBE has quit [Ping timeout: 264 seconds]
maartenBE has joined #scopehal
<Degi>
Can I use 0.25 mm traces on 0.35 mm pads?
<azonenberg>
probably ok, but why use something so huge?
<azonenberg>
the dogbone trace is super short, R/L are tiny for the most part
<Degi>
Idk less inductance?
<Degi>
Okay
<Degi>
Wait nvm this is not gonna be high speed, max 500 MHz
<Degi>
Is 0.15 mm ok or should I use like 0.127?
<azonenberg>
I generally make dogbone traces whatever my default trace size is
<azonenberg>
unless i have a reason to do something else
<azonenberg>
the big thing you want to avoid is having a mix of different sizes with a few fat ones that will potentially cause thermal nonuniformities during reflow
<azonenberg>
this is the z-ground installed on the probe
<azonenberg>
by rotating it, you can vary the signal to ground spacing
<electronic_eel>
azonenberg: thanks
<electronic_eel>
but you also have to rotate the probe then, right? this means you need a more flexible coax cable
<electronic_eel>
currently I use LMR-195 UF on the bench, I may have to use something thinner and more flexible for the probe then
<azonenberg>
So thats one of the things holding me back right now. in particular the bipod positioner only holds the probe at one orientation
<azonenberg>
which limits how much the z-ground can move
<azonenberg>
(because the probe is too fat to fit in the bipod any orientation other than edgewise)
<azonenberg>
if you have ideas on an alternative positioner, perhaps one full custom 3d printed or similar, that might open a lot of doors on this front
<electronic_eel>
can you show photo of the back of your tetris in the holder? I'm not sure how they exactly line up
<electronic_eel>
maybe also of your current probe prototype with it's case in the holder
<miek>
i guess their solution for that is the noga-arm & clamp style of positioner
<azonenberg>
So this is *very* interesting
<azonenberg>
the lecroy positioner and the pico positioner are NOT the same
<azonenberg>
They're made on the same injection mold, but the zs1500 / tetris positioner appears to have had the top gap machined out slightly to accommodate the fatter body of the tetris
<azonenberg>
the opening is about 800 μm wider and there's no mold flash
<azonenberg>
it's smooth, obviously was done post injection molding
<azonenberg>
(tool marks elsewhere on the mold ejector pins etc confirm they're made on the same die)
<electronic_eel>
can you rotate the probe in the lecroy positioner?
<azonenberg>
so here's a "standard" scope probe with a tapered boy
<azonenberg>
body*
<azonenberg>
https://www.antikernel.net/temp/IMG_20200326_122938.jpg here's my current probe. Which is going to be getting taller in order to accommodate the wider ground-to-signal spacing i need for tetris ground compatibility
<electronic_eel>
doesn't have pico also rebranded tetris probes?
<azonenberg>
I havent looked at pico's active probe offerings
<azonenberg>
would not surprise me, since we know pico sells some rebranded pmk products
<miek>
they have 893-250-001, 893-250-4SO (sonic), and 893-250-00T (tetris)
<azonenberg>
893-250-001 is what the pico blue one likely is
<azonenberg>
the lecroy black one is likely a custom color version of 893-250-00T
<azonenberg>
and lecroy sels 893-250-4SO with the zs4000
<azonenberg>
i dont have a sonic/zs4000 in my lab though
<azonenberg>
i intend for my probe to fit 893-250-001 as i can source the pico branded version of it cheaply from tequipment for much less than STI sells the PMK branded one
<electronic_eel>
oh, so a whole lot of slightly different positioners out there
<azonenberg>
Yes. The 250-4SO is likely a different injection mold, but i dont have one to confirm
<azonenberg>
Not possible, not even close. the width of the slit in the positioner is about 4 mm, minimum plastic wall thickness for most 3d printing and injection molding processes is around 1 mm
<azonenberg>
so that leaves 2mm for the pcb which is just not going to happen
<azonenberg>
With a full custom positioner that has a bigger bore? totally
<azonenberg>
but that's a lot more engineering
<azonenberg>
my current probe is 5mm wide PCB, and the OD of the shell is 9.5 mm in the fat area and 7 mm in the shaft
<azonenberg>
so literally as thin as i can make it
<electronic_eel>
what is the diameter of the hole in the positioner, below the slit?
<azonenberg>
way too small, didn't measure
<azonenberg>
probably 4mm or so judging by eye
<azonenberg>
The probe pcb is going to actually be getting wider, closer to 1 cm in the shaft area, in order to accommodate the increased ground-to-signal spacing of the tetris ground socket
<azonenberg>
so i'll be changing the shell geometry from cylindrical to more of a rounded rectangle shape
<azonenberg>
keeping the skinny axis small enough to fit in the positioner edgewise
<azonenberg>
Making it arbitrarily rotateable would require a completely different positioner design
<azonenberg>
Not that i'm opposed to doing that, but it would be more work
<electronic_eel>
ok, that is not going into the positioner and be rotatable...
<azonenberg>
exactly
<miek>
does it need to be that wide all the way down? could it thin down towards the cable?
<azonenberg>
miek: It could but then you'd have a really weird looking shape with a bulge down towards the tip
<azonenberg>
for ergonomic/visual reasons i'd rather keep it fat at one end and slowly neck down towards the tip
<azonenberg>
i don't waant it to look like an exaggerated boeing 747 with a big hump on the nose
<electronic_eel>
how about this: a 3d printed socket that fits into the slit+hole of the pico positioner. on top of this socket sits another, larger hole which holds the actual probe
<azonenberg>
at that point i'd rather just build my own positioner from scratch
<azonenberg>
3d printed plastic shell, put some self tapping steel lag bolts into the feet to add mass, then some rubber feet of some sort
<azonenberg>
same basic shape but sized for the new probe design
<azonenberg>
oh wow that was fast. tequipment quoted $22.20 @ qty 100 for the pico ta102 with a 3-4 week lead time
<electronic_eel>
I heard there are companies out there that do injection molds with aluminium tools which work for a few hundred pieces
<azonenberg>
For beta units i'll be doing 3d printed from shapeways. i have test versions of the current shell out at them in a couple of different materials
<electronic_eel>
maybe that would be an option for a run in the lowish hundred pieces size
<azonenberg>
all nylon based, some SLS and some MJF, with and without glass filling
<electronic_eel>
for beta 3d prints of course
<azonenberg>
the current shell in SLS nylon, no glass fill, is only $5. not sure how much more a positioner would be
<azonenberg>
But it might actually be cost effective to 3d print
<electronic_eel>
it is probably easier to manage, as you don't have to care much about the production process
<azonenberg>
i'm very much not a fan of FDM, i hate it, and i think it has no place in production or even serious prototyping
<azonenberg>
but i love SLS
<azonenberg>
I just cant afford/justify having the equipment in house, so i outsource
<azonenberg>
anyway, that's kinda the roadmap. i want to get back to the psu now
<azonenberg>
Related: I plan to make a hmcad1520 testbed as an INTEGRALSTICK host board in the next few weeks. it will be intended for prototyping in combination with the afe characterization board to make a single channel scope
<azonenberg>
just because of all the lead time, i want to have it in the pipeline while i wait for other stuff
bvernoux has quit [Quit: Leaving]
_whitelogger has joined #scopehal
_whitelogger has joined #scopehal
<electronic_eel>
lgtm
<electronic_eel>
not exactly the cheapest, but reliable
<Degi>
But nice ESR
<electronic_eel>
esr in this range is to be expected for a good polymer cap
_whitelogger has joined #scopehal
<Degi>
What are the i2c/spi voltage levels?
<azonenberg>
how high do you think this cap is?
<Degi>
A few mm?
<azonenberg>
mostly 5V because the analog subsystem is running at 5V
<Degi>
Okay
<electronic_eel>
10.3mm according to digikey
<azonenberg>
so why is that a problem? it's going to be in a 1U case, or a desktop form factor almost the same height
<electronic_eel>
you won't be able to stack smas that tightly
<Degi>
Hm right
<azonenberg>
in the entry level system the ADC+4x AFE boards will be horizontal anyway
<azonenberg>
i don't plan to go vertical until higher end systems
<azonenberg>
and the details of form factor for the bigger systems are TBD. one possible option would be eurocard with pcie connectors to the backplane, similar to my MARBLEWALRUS concept that never really got implemented
<azonenberg>
in which case each card would be 3U high x 160mm deep x 4 HP (~20 mm) wide
<Degi>
Lol for maximum parts density, watercooled scope
<azonenberg>
lol no
<electronic_eel>
azonenberg: how about 565-APXJ160ARA221MF61GCT-ND instead? costs only half of the other
<electronic_eel>
16v should be enough for a 12v input
<azonenberg>
i like derating more than that
<Degi>
And hts smaller
<azonenberg>
for electrolytics my rule has always been ~2x
<electronic_eel>
for classic electrolytics yes, but for polymers you can go tighter
<azonenberg>
ok that will work then
<Degi>
And lower ESR
<azonenberg>
Lower ESR in this case actually has limits
<azonenberg>
some regulators will destabilize if it gets too low
<Degi>
Hm isn't that for the 12 V rail?
<azonenberg>
they explicitly say no lower than 17 in the datasheet
<electronic_eel>
but not on the input
<azonenberg>
electronic_eel: i'm using the same cap on the -7V output
<azonenberg>
they said min 100 uF and more is better
<azonenberg>
negative rail is using the PTN78000AAZ
<electronic_eel>
ah, ok, for the output the esr is a thing
<Degi>
Hmm my ECP5 dev board doesn't use any electrolytics at all... But I guess this could benefit from a more ripple free rail.
<azonenberg>
Degi: what voltage does it run at?
<Degi>
3.3 V and 12 V input
<azonenberg>
i normally use all ceramic up to 5V but at 12V if you need a lot of capacitance, it gets hard
<Degi>
There are some 10 µF 25 V 1206 caps
<azonenberg>
yeah but 200 uF?
<azonenberg>
also if you have a min esr constraint that gets hard
<Degi>
Hm yeh
<electronic_eel>
10 of these 1206 caps? you only do that if you are height constrained or something similar
<azonenberg>
also i bet those caps are... maybe 3 uF @ 12V
<azonenberg>
look at the c/v curves. They're brutal
<electronic_eel>
the min esr can be "fixed" by putting a resistor in series
<azonenberg>
3 uF i'm being optimistic
<electronic_eel>
but at that point you are better off looking for some other solution
<azonenberg>
electronic_eel: i used a 380 uF 1812 mlcc on a fpga core rail once
<azonenberg>
but that was a 1V rail
<electronic_eel>
1812? these get extremely sensitive to cracking
<electronic_eel>
and if you have a beefy psu they will then burn your whole board
<electronic_eel>
(or house if you you have bad luck)
<Degi>
Put them on the underside, maybe they desolder themself and fall off
<electronic_eel>
yeah, and then a imu unit to switch off the board if it is installed in the wrong orientation
<electronic_eel>
I actually had a problem with inductors desoldering themselves and falling off boards several years ago
<Degi>
Lol
<electronic_eel>
it was a switcher regulated in software on a mcu and sometimes the regulation speed of the switcher and the load pulses created by the same mcu lined up in a bad way
<azonenberg>
Lol i heard a story about somebody who had a spring loaded heatsink that had a similar effect
<Degi>
Lol
<azonenberg>
when a fet got overloaded, it would self desolder and pull off the board
<azonenberg>
boom thermal shutdown :p
<Degi>
As long as that doesnt make an arc fault lol
<azonenberg>
you want scary power supplies look at crypto miners
<electronic_eel>
yeah, crypto miners have crazy hardware
<azonenberg>
i saw pics of one under development that uses a ton of intel cpu vrms in a massive multiphase dc-dc
<Degi>
Hm I found out that those big TO packages can work as flamethrowers too
<azonenberg>
the output is 900 mV at about 1.5 kA
<Degi>
Lol
<Degi>
To power the asics?
<azonenberg>
Yeah
<azonenberg>
like, kW class PC power supplies put out 12V at high tens to low hundreds of amps. Which is... less insane
<azonenberg>
but sub-1V kW class supplies are NUTS
<electronic_eel>
azonenberg: why did you chose to use the isolating dc/dc for negative and not the buck solution I proposed?
<azonenberg>
electronic_eel: it's non isolating, no?
<azonenberg>
same ground on both sides
<electronic_eel>
ah, sorry, misread
<azonenberg>
I figured a prebuilt module would be a low effort, low risk solution
<azonenberg>
actually would have used one for the +7 rail except digikey's website was derping last night and the category for board mount dc-dc modules wouldn't load
<electronic_eel>
but a bit pricey. a buck may be 2 or 3 bucks
<azonenberg>
We can look at alternative options during cost-down once we have everything working
<azonenberg>
my key goal right now is getting the AFE tested and debugged
<azonenberg>
and i want to do it in as few board spins as possible in these uncertain times for the supply chain
<electronic_eel>
idea: we aren't size constrained on the eval board. how about putting both on the board and if the cheaper solution doesn't work you can populate the module
<azonenberg>
The eval board is going to be made at oshpark which charges per unit area
<azonenberg>
so... :p
<electronic_eel>
ok
<Degi>
(Idk JLCPCB charges the same below 100x100 mm)
<sorear>
I wonder if running a few dozen miners in *series* would be useful
<Degi>
Well as long as they use exactly the same current
<azonenberg>
oh so we still have the overvoltage_n going negative issue to deal with