futarisIRCcloud has quit [Quit: Connection closed for inactivity]
<Famine>
whats the allowed error tolerance on the torque ?
<azonenberg>
Famine: Good question. For brass SMAs I see 3-5 in-lbs as the typical range
<azonenberg>
but i assume for repeatability you want better than 4 +/- 1
<azonenberg>
(that's 0.34 - 0.57 NM)
<azonenberg>
i see a fairly expensive one from Fairview Microwave, in the screwdriver form factor i like, that says +/- 4%
<azonenberg>
set for the upper end 5 in-lb / 0.57 Nm)
<azonenberg>
i've been meaning to buy one but have had higher priorities
<azonenberg>
and to date, none of the stuff i've been doing has seen problems from being hand tightened to a "reasonable" level
<azonenberg>
So it hasn't been a priority
<Famine>
that's not bad accuracy, i just checked wiha and the best they do is +-6%
<azonenberg>
this is also a $429.89 torque driver
<azonenberg>
so... :p
<Famine>
wiha's was ~$400
<azonenberg>
meanwhile the torque driver i have in my toolbox is +/- 6% adjustable from 8-40 pounds
<azonenberg>
but was $60
<azonenberg>
in-lb*
<azonenberg>
but i also use it for mains power terminals, which are... slightly less critical tolerances than SMAs
<Famine>
tohnichi makes a really cool one +-1% but its ~$900
<azonenberg>
i'm kinda curious how close i can get finger tightening
<azonenberg>
with a bit of practice
<azonenberg>
but i dont have a torque sensor i could test on
<Famine>
could always rig up something janky with a load cell lol
<azonenberg>
Lol
<Famine>
sometimes the ##electronics people make me want to stab them with a rusty spoon.... i was talking about some low noise high current supplies and every second comment was "why linear? why not use xyz random aliexpess SMPS"
<Famine>
but anyways, i'm working on a new power supply or possibly a space heater given all the losses.
<monochroma>
Famine: lol, that's pretty much why i stopped hanging out there
<Famine>
monochroma, i don't blame you, its a decent enough channel for people happy just doing low voltage digital with arduino but not much else
<azonenberg>
yeah i idle in there but dont usually actually talk unless summoned for some reason
<azonenberg>
i like places where i can talk about microwave and RF and actually get useful answers :p
<monochroma>
azonenberg: i think that's where we met originally :o
<azonenberg>
quite possibly
<azonenberg>
i mean i introduced you and lain on irc :p i forget what chans she hung out in at the time though. maybe ##fpga?
<Famine>
here is a fun problem, i need to absolutely prevent any reverse transients >10uA on a 100A CC supply x.x
<azonenberg>
Famine: o_O
<azonenberg>
10 uA on 100A?
<azonenberg>
that's 10^-7
<Famine>
azonenberg, well max spec is 25uA on 40A, but i want to run 2 modules off the same supply so i figured 1/2 max spec
<Famine>
would be a good point to shoot for because i know i'll probably blow my error budget quickly
<azonenberg>
yeah
<Famine>
Negative current transients greater than 25 µA, and/or reverse voltages >3V, can destroy the unit < i always pick fun projects
<lain>
D:
<Famine>
on the bright side my laser was dirt cheap, now i just need to spend a grand on a power supply, optics, and goggles
<lain>
definitely goggles
<lain>
speaking of which, do you know of a good way to verify that goggles are actually working properly? I have a pair that seems legit, but I have yet to come up with a suitable test to verify that
<Famine>
lain, i'm looking at some LG9C's from thorslab
<Famine>
lain, i don't so i'm just grabbing a certified pair
<azonenberg>
lain: if you have any doubts i'd say throw them out and buy ones from a reputable supplier
<azonenberg>
like thorlabs or edmund optics
<azonenberg>
that's where i got my uv glasses i use for uv-cure resins etc
<Famine>
i plan on using my eyeballs for a while yet, so $200 is cheap insurance
<azonenberg>
That seems like a reputable source, what power level / wavelength are you using?
<azonenberg>
never heard of them but it doesnt look like a sketchy aliexpress seller :p
<lain>
yeah
<Famine>
they have certs listed for everything so they seem pretty legit
<lain>
the source is a Tria Precision laser hair removal device. internally it uses a 40W diode bar laser, and it's pretty safe by itself -- the aperture is a diffuser, because they found that skin is a diffuser at the optimal hair removal wavelength, and stacking two diffusers makes no appreciable change to the effectiveness of the laser for hair removal
<azonenberg>
oh that terrifying monster
<azonenberg>
lol
<lain>
it's only pulsed briefly, and will only activate when it contacts skin, and specifically only skin light enough to be safe to use with it
<lain>
but it's still a frightening laser
<azonenberg>
So you're saying your laser is a racist?
<lain>
if you're not wearing any sort of magnifying glasses or other optics, the diffuse output is eye safe, at least as long as it's about a foot away
<lain>
but if you're using it on your face, that's another matter
<lain>
also just in general I don't trust 40W anything when eyes are involved
<azonenberg>
i assume the small ones are aiming telescopes and rangefinders etc
<monochroma>
2 tiny ones, 2 medium size ones, and then the mega chonker
<azonenberg>
yeah
<azonenberg>
and all of them have laser warning markers over them
<Famine>
lain, only 30w of 808nm with a 800um beam width at the output x.x
<Famine>
it's about the right mix of excitement and pants peeing terror for me
<lain>
:D
<lain>
we got a "broken" Tria Precision device to inspect, they go for dirt cheap on ebay
<lain>
haven't had time to try to dump the firmware, but it seems the battery came disconnected or otherwise failed, and it caused the device to flag an error. it no longer functions, instead just always reporting an error during startup
<lain>
however, you can power the laser diode directly, and it does appear to light
<azonenberg>
i personally prefer to avoid anything in "do not look at beam with remaining eye" territory
<Famine>
lain, the crazy part about the tria is its not completely potted (just looked at the tear downs)
<lain>
we tested this by uh... setting up the bench psu with no laser attached, then unplugging the psu from mains, making sure it's discharged, making sure it's discharged, attaching it to laser, then setting up an old camera to record the laser, leaving the room completely and closing the door behind us, going __around the corner behind another closed door__, then plugging in the psu from there :P
<lain>
Famine: yeah
<lain>
there's terrifying hack potential :P
<lain>
my main interest is in why the busted units are busted. it's either that some transient failure occurred which has since resolved, and the mcu blew an error flag for safety... or it reached some sort of diode lifetime limit and blew an mcu fuse to signal that
<lain>
I'm curious if the diodes are still usable in most of the "dead" units
<lain>
because if so, that'd be a much cheaper way to get one
<Famine>
i mean even if the diode array is even close to the rated power it could make for some interesting projects
<Famine>
azonenberg, thats a good way to avoid adding another expensive hobby to the list lol
<azonenberg>
lol
<azonenberg>
building oscilloscopes is expensive enough :p
<Famine>
yep, i think i'm looking at ~1.5k all in for my low power laser engraver
<Famine>
but it's starting to spawn crazy ideas .... https://i.imgur.com/0lKAKBb.jpg only $1k and a decent starting point for a DPSS q switched YAG
<lain>
wuzzat do :3
* lain
doesn't actually know much about lasers, just happens to have a 40W hair zapper :P
* monochroma
has a pile of what she thinks are CO2 lasers from the mid to late 80s that clearly came off a large optical bench, likely from a DoD lab
<Famine>
lain, its basically a huge array of bar diodes, in a reflective cavity. so they put a neodymium YAG rod in the center, and then use the diodes to pump energy into it until it starts lasing
<monochroma>
yo dawg, i heard you liked lasers...
<Famine>
lain, basically think of the laser diodes as a power supply, then the YAG rod as a cap, and the Q switch as a transistor. you can dump a huge amount of power into the rod then dump it all at once for crazy high peak powers
<azonenberg>
purple is the reference plane for the green signal layer
<azonenberg>
i'm trying to keep as much as i can on the front layer which has a solid ground reference, but that isn't practical with the constraints of this connector
<azonenberg>
in retrospect i should probably have had the lvds all on one side when i designed integralstick but it's too late for that now
<azonenberg>
I don't care about the low speed stuff but trying to keep a clean ref plane for the LVDS and analog signals
<azonenberg>
as far as the back goes, LCLK is the long signal going up to the far end of the connector
<azonenberg>
it's also the reason for all of the squiggles elsewhere, because LCLK is much longer than all of the data and i'm trying to phase match :p
<azonenberg>
most LVDS on the back side is referenced to the 1v8 pour, but LCLK crosses over 3v3
<azonenberg>
then the fpga-side LVDS is referenced to 2v5 so i will need 1v8 - 2v5 caps near the q-strip
<azonenberg>
LCLK already has a 3v3-1v8 stitching cap
<azonenberg>
then data channel 1A crosses over no less than FIVE reference plane splits in the current routing :p
<azonenberg>
So much for being almost done
<Famine>
azonenberg, you need work to buy you an altium seat :P
<azonenberg>
why?
<azonenberg>
first off, windows eew. Second, $dayjob i barely do any board design so there's no way they'd spring for it
<azonenberg>
third, what does it do that kicad can't for this sort of stuff?
<Famine>
with the new update they made length matching / diff pair routing almost too easy
<azonenberg>
now that i have diffpair routing and length matching i'm pretty happy. the length matcher does need some fine tuning especially for diffpairs, but mostly i've been very happy with it
<monochroma>
kicad is pretty rough around the edges, a lot of commercial CAD packages are a /LOT/ more efficient at PCB layout, part entry, schematic capture etc. it's mostly $PEOPLE got paid to make making boards quick and efficient. (and better DRC and such)
<Famine>
monochroma, i've been spoiled with a lot of commercial software for free. went back to school and found out i could whore myself out to big software companies
<monochroma>
lol
<azonenberg>
yeah - i prefer to use free tools when possible but if they don't exist or are vastly worse (e.g. openems vs sonnet) i'll buy the commercial tool if i can justify it
<azonenberg>
so far, kicad hasn't been problematic enough for me to want to buy a commercial eda tool to replace it
<monochroma>
i wouldn't mind commecial software as much if their derpy DRM systems on paid for features/software didn't keep getting in the way of work/the software didn't cost $$$$ for continued support etc
<Famine>
i've been trying to get into octave, only 1 really annoying bug so far and its windows 10 not octave
<azonenberg>
i was never into matlab in the first place
<azonenberg>
so that never was an issue
<Famine>
the amazingly bad scaling feature in windows 10 requires me to start octave, drag the window to my secondary monitor (larger resolution) then back to my main monitor to update the font scaling
<monochroma>
lol yay derpy DPI issues
* monochroma
closes all of her ebay tabs for the shiny shiny Yag rods >_>
<Famine>
monochroma, haha it's an addiction
<monochroma>
i'm kinda grumpy, before i knew what they were, goldmine-elec.com used to sell yag rods for cheap for YEARS, but by the time i knew what they were, they were gone :P
<Famine>
if you are in the US you can probably score a SSY-1 for a decent price
<monochroma>
these were like $12
<_whitenotifier-3>
[starshipraider] azonenberg pushed 1 commit to master [+0/-0/±5] https://git.io/Jvbzg
<_whitenotifier-3>
[starshipraider] azonenberg ecbce5d - Continued layout on ADC test board. Fixed a bunch of plane crossings.
<azonenberg>
How close would you consider minimum spacing for that?
<electronic_eel>
hmm, let me try
<electronic_eel>
20mm works well, 15mm won't work
<azonenberg>
So for the sake of discussion, a 1U enclosure is about 415 mm wide
<electronic_eel>
would have to jig it somehow to get a exact number
<azonenberg>
not counting ears
<azonenberg>
suppose i have 130 mm on the left side marked out for the display that we may or may not end up using, but i want to budget for that
<azonenberg>
that leaves 285 mm for eight input connectors and external trigger, so nine. very roughly that's a max of 31 mm
<azonenberg>
i'm targeting 25 mm width for each AFE to provide room for spacing between boards, emi cans over the afes, etc
<azonenberg>
So should have plenty of room to spare
<electronic_eel>
yes, looks good
<electronic_eel>
also plan a row of usb-cs for the logic analyzer inputs
<electronic_eel>
hmm, would a column be the better word?
<azonenberg>
My plan is to have everything be SMT on the AFE board
<azonenberg>
with the usb-c's between the SMAs
<azonenberg>
you can remove the LA cable to torque the sma if needed
<azonenberg>
s/LA/active probe/
<electronic_eel>
yes, the usb-c for the active probes are between the smas, or maybe on top
<azonenberg>
between. top would mean some kind of cable off the main board
<electronic_eel>
but the ones for the LA, with full ss data, need some space too
<azonenberg>
between we could do with smt
<azonenberg>
LA will be in the same area as external trigger
<azonenberg>
on the far right side of the scope, i'm thinking the digital board will be L-shaped
<electronic_eel>
you could use a separate board just for the probe usbs and stack that on top
<azonenberg>
Yeah we'll have to tinker a bit and see what makes sense
<azonenberg>
i'll probably do some rough budgeting in solvespace
<azonenberg>
i like using constraint solvers to play around with mechanical design concepts
<azonenberg>
actually... trigger board has to be separate
<azonenberg>
ext trig
<azonenberg>
we can put LA there too
<azonenberg>
but basically, i don't want to deal with the mechanical issues of trying to put edge launch connectors on both sides of a board that fills the entire footprint of the case
<azonenberg>
how do you get it INTO the case?
<azonenberg>
if you only have connectors on one side you can go down and sideways
<azonenberg>
so if we have a q-strip for each of the two analog boards and one for the trigger/la board, that would makes sense
<azonenberg>
we also have to figure out the design of the LA itself
<azonenberg>
Might want to make a standalone prototype of that too
<electronic_eel>
yeah, the la design has to be at least sketched up before you begin building the connection for it
<azonenberg>
tentative thought is to have an active probe head containing an artix-7 and probably two 8-bit comparator based input stages
<azonenberg>
actually, maybe we dont even need an FPGA in the head
<azonenberg>
oh wait we do if we use usb-c
<azonenberg>
because we have to serialize all of the comparator output data out into a single diffpair for data
<electronic_eel>
yes, you need to serialize the data to get it through usb
<azonenberg>
we need to figure out the alt function design etc as well
<azonenberg>
what pins we use for what
<electronic_eel>
the other thing is proper timing, you want the clocks of the main scope fpga and the one on the la to be in sync
<electronic_eel>
and determine the phase lag between them
<azonenberg>
Yes. Tentative plan is one of the diffpairs to be refclk from host to device
<azonenberg>
from main to la
<azonenberg>
one be trigger from main to la, one be trigger from la to main
<electronic_eel>
yes, that should work
<azonenberg>
and one be waveform data
<monochroma>
could just use a fast shift register and a clock on the probe end ;)
<azonenberg>
as far as determining the phase lag, that will be fun because the cable could vary in length
<azonenberg>
we can probably do something with the trigger out/in signals to calibrate
<azonenberg>
assuming round trip each way is the same
<azonenberg>
i feel like this will have to be a self cal of some sort
<electronic_eel>
you mean the different pairs within the usb-c cable have different lengths?
<azonenberg>
no
<azonenberg>
i mean the cable can be replaced by the end user
<electronic_eel>
yes, of course
<azonenberg>
we aren't cal'ing at the factory to a specific cable
<electronic_eel>
yes, this would be annoying for the user
<azonenberg>
So... let me think a bit
<electronic_eel>
so there should be some kind of self cal for the la
<azonenberg>
oh, derp
<azonenberg>
probes use SMA cables that can vary in length too
<azonenberg>
we only care about skew from probe to LA
<azonenberg>
so what if we just have a cal you can trigger on request that involves you touching a scope probe and LA channel to any kind of repeating waveform?
<electronic_eel>
I think that is reasonable
<electronic_eel>
usually the scopes have a cal output for this kind of thing
<electronic_eel>
also for the classic passive probes of course
<azonenberg>
well normally the output is for compensating 1M passive probes
<azonenberg>
:p
<azonenberg>
but yes we can probably make some kind of cal output for this purpose
<electronic_eel>
but couldn't we send a pulse down one of the trigger pairs and receive it back on the other?
<electronic_eel>
I mean the usb to the la
<azonenberg>
yes. but that won't tell us anything about the probe cable length on the analog side
<azonenberg>
we care about a channel to d channel skew
<azonenberg>
We may also want to allow channel to channel deskew, to compensate for propagation delays of the cables if you dont use the same length sma cables on each channel
<electronic_eel>
yeah, you have the a probe to scope cable and the la probe clip/wire to la pod cable too
<azonenberg>
so thats why i'm saying we don't actually care about any individual delay
<azonenberg>
all we care about is deltas
<azonenberg>
absolute value is of no consequence, right?
<electronic_eel>
shouldn't go off the buffer of the scope, but otherwise I don't see an issue
<azonenberg>
we can certainly measure trigger propagation separately, since we do want the la and analog channels to trigger at the same time
<electronic_eel>
but I'd say the scope and la are constantly sampling. "trigger" is something that is defined at a later time and always comes with a timestamp how long ago it was
<azonenberg>
yes, we'll capture to a ring buffer and then stop X time after a trigger
<azonenberg>
we may want a separate buffer on the LA head
<azonenberg>
because i dont think we'll have bandwidth on one diffpair for all of the LA channels to stream in real time
<azonenberg>
we'll need to trigger then download to the scope fpga
<azonenberg>
let's see, what's a reasonable sample rate for digital channels on a 100 MHz scope?
<electronic_eel>
ugh. I don't like that. because the la pod will then have to become big to house SO-DIMMs
<azonenberg>
no we could probably use something smaller since it isnt a HUGE amount of data
<electronic_eel>
also you can't do intelligent cross triggering between analog and la
<azonenberg>
well let's do the math then
<azonenberg>
in a -2 artix7 we can do 1.25 Gbps on a DDR LVDS RX
<azonenberg>
suppose we do 1 Gsps to keep the numbers round and align sample rate to the main scope
<azonenberg>
times sixteen input channels (two 8 bit banks) is 16 Gbps
<azonenberg>
6.25 Gbps max raw line rate for non flip chip artix7 GTPs
<electronic_eel>
that isn't something for one usb-c pair...
<azonenberg>
after 8b10b we only have 5 Gbps of application layer throughput
<azonenberg>
sure, we could cut down the sample rate, but at that point we're handicapping ourselves
<electronic_eel>
yes, 1 gsps sounds reasonable to me to make it useful
<electronic_eel>
maybe rethink the usb-c idea for the la?
<azonenberg>
If we used a cable that had four diffpairs, we could do 20 Gbps
<azonenberg>
which is enough for realtime streaming
<azonenberg>
but that's four data pairs plus refclk, trigger in, trigger out is at least seven pairs
<electronic_eel>
for connecting lot's of hard disks and ssds in server, there are several sff cable standards, we could use one of those
<monochroma>
USB-C cables have 2 diffpairs for SuperSpeed mode
<azonenberg>
yes but do you know how pricey those cables probably are?
<monochroma>
you can also get special USB-C cables that have 4 diffpairs (using the top and the bottom pins)
<azonenberg>
monochroma: is that not standard?
<azonenberg>
i thought everything had four pairs
<lain>
the connectors have 4 pairs
<azonenberg>
and then figured out which ones did what when you flipped it
<lain>
the cables have 2
<azonenberg>
... oh
<monochroma>
they have an internal mux iirc
<azonenberg>
there goes that idea before we started. So usb-c is an absolute no for the LA
<lain>
yeah they use muxes at each end to choose the appropriate pair based on USB CC negotiation
<azonenberg>
so all usb c cables are active?
<lain>
which is also how they can just not mux anything onto those lines, or mux analog stuff onto them, if it's not a usb cable
<lain>
no
<lain>
that's in the host/device
<azonenberg>
i thought they had muxing on the host and device for four pairs in the cable
<lain>
they only use 2 pairs for usb
<lain>
things like displayport alt mode use 4 pairs
<azonenberg>
so you can't put displayport alt mode on the far side of a standard usbc cable?
<lain>
I'm... not sure
<lain>
checking the standards now
<azonenberg>
electronic_eel: sooo, we can cut down the pin count a bit
<azonenberg>
actually no nvm
<azonenberg>
we don't have a high speed return link
<azonenberg>
i was gonna say use the CDR recovered clock
<electronic_eel>
how about SFF-8087? it has 8 diffpairs+sideband
<azonenberg>
but we only have the fast serial one way
<monochroma>
yeah the server SAS connectors arn't too bad
<azonenberg>
So we could nix the whole usb idea for the LA then, use 4 diffpairs for LA return data streaming to the host
<azonenberg>
one for refclk from host to device, one for trigger from host to device, one for trigger from device to host
<electronic_eel>
now with 8 pairs, you could also do a fpga-less la for 8 channels
<azonenberg>
and then use the last pair would be a return GTP link
<lain>
well heck
<lain>
I stand corrected
<lain>
it seems like cables may actually have all 4 pairs
<azonenberg>
so we can send config commands to the fpga via 8b10b coded data at 5 Gbps
<monochroma>
hmmm
<electronic_eel>
lain: I think this depends on the exact cable type. there are like 7 different cable types out there...
<azonenberg>
So one bidir GTP, three unidirectional GTPs, refclk, bidir trigger
<azonenberg>
then use sideband for power
<lain>
electronic_eel: so there's usb-c to not-usb-c cables which will only have exactly what they need (e.g. usb-c to usb-a only needs 2 pairs, and may not even support usb3 at all in which case it lacks all the fast pairs and only has usb2 pairs)
<lain>
but for usb-c to usb-c, at least according to the spec that everyone constantly violates, it looks like they should have all 4 pairs populated for exactly the reason that you might plug a full 4-pair-using device into another 4-pair-using device
<electronic_eel>
azonenberg: why not put just the comparators into the la pod and transmit the diff output of each comparator through a separate pair?
<lain>
and I would tend to believe that, because thunderbolt is also using usb-c cables and, afaik, is using all 4 pairs. also displayport alt mode and displayport+usb3 alt mode use all 4 pairs, and should work with a generic c-to-c cable
<azonenberg>
electronic_eel: because i want two 16-bit banks, which is pretty much the standard that every scope-attached LA has
<azonenberg>
two 8-bit*
<azonenberg>
that's 16 signals
<azonenberg>
we'd need two sff-8087 cables
<electronic_eel>
but why not just have two la pods for that?
<azonenberg>
i guess we could have two separate LA pods
<electronic_eel>
would make the la much simpler
<azonenberg>
yeah, no FPGA at all
<electronic_eel>
and also more flexible, as a faster fpga in the scope could be leveraged for the la too
<azonenberg>
just a bank of comparators and a spi dac or something (via sideband) for setting Vt
<electronic_eel>
exactly
<azonenberg>
Ok, sounds like we're settled on that then
<azonenberg>
exact pinout assignment to come later
<azonenberg>
but probably 4 sideband for spi control and 4 for power
<electronic_eel>
or i2c dac?
<azonenberg>
i dont trust i2c as much as spi over long cables
<electronic_eel>
could also do detection
<monochroma>
also note micro/mini HDMI or mini display cables are also options. they tend to be much thinner and more flexible than the SAS connectors
<electronic_eel>
rs232 and a small mcu in the la pod?
<azonenberg>
electronic_eel: that could work too
<azonenberg>
ttl level serial not rs232
<azonenberg>
(that way we can presence detect too, ping the mcu)
<electronic_eel>
ok, ttl level is fine by me
<lain>
ok here we go: specifically usb 3.1 cables are what will have all 4 pairs for c-to-c. this is also called "full featured" but the key here is 3.1, not 3.0
<azonenberg>
monochroma: mini hdmi cables only have 4 diffpairs
<azonenberg>
sas have 8
<monochroma>
azonenberg: keep in mind the SAS cables are usually very stiff and usually come in shorter lengths
<azonenberg>
i see a $12.99 one on amazon that's 1m long and claims to be flexible
<azonenberg>
at that price i might buy it just to see
<azonenberg>
plus a dac and mcu of negligible current consumption
<electronic_eel>
monochroma: that looks stiff
<monochroma>
electronic_eel: yup :P
<monochroma>
NXP uses them for MIPI for... some... reason...
<electronic_eel>
the ones I have in the servers are much nicer
<electronic_eel>
don't know the exact brand offhand tough
<azonenberg>
electronic_eel: at 5V Vdd, it uses 12 mA VCCI, 25 mA VCCO per channel. So 37 mA * 8 channels = 296 mA * 5V = 1.48W, adding in conversion losses plus mcu/dac let's round up to 2W
<monochroma>
i think these will always be a bit stiff for side to side movement, compared to a round cable but should work
<azonenberg>
at 12V if we feed the base 12V from the wallwart out to the LA, that's 166 mA
<azonenberg>
over six conductors is 28 mA per conductor
<electronic_eel>
yeah, with 12v this seems to be easy
<azonenberg>
30awg can handle up to 500 mA without overheating, but let me see about voltage drop
<electronic_eel>
we just need some detection scheme that we don't fry some other accidently connected gear with the 12v
<electronic_eel>
like some pullup/pulldown resistor combination
<azonenberg>
103 ohms per 1000 feet, so for a 1 meter cable that's only 0.337 ohms
<azonenberg>
at 28 mA that's 10 mV of drop
<electronic_eel>
make it 0.5 ohms with the connectors
<azonenberg>
ok 14 mV
<azonenberg>
over 12V that's insignificant
<azonenberg>
maybe we use five of the sidebands instead of six and the last as a presence detect?
<azonenberg>
we clearly dont need all six for power based on my math
<azonenberg>
have some magic resistor value there to ground, apply 3.3V via 10K and see what we get on the other side
<azonenberg>
we already need an adc host side anyway
<electronic_eel>
we could use the uart ones for detection, specific pullups there
<azonenberg>
then you are fighting whatever the mcu does when powered down
<azonenberg>
an unused pin is guaranteed to not have anything else on it
<azonenberg>
we can afford the additional resistance
<electronic_eel>
ok, in this case we really have the pin to spare
<azonenberg>
heck, even at 3 pins we're only at 55 mA per conductor, rated at 500 mA thermally
<azonenberg>
and voltage drop will double with 2x the resistance to 28 mV out of 12V
<azonenberg>
Or 0.2% power lost in the cable
<azonenberg>
So i guess what makes sense at this point is, i'll finish the final design review on the adc test board and order that in the next couple of horus
<azonenberg>
then start doing actual design work on the LA head
<azonenberg>
we can build the LA head now, it doesn't depend on anything else, then make another integralstick module that talks to it for prototyping
<azonenberg>
actually no we can't do the host
<azonenberg>
oh wait we're not using GTPs
<azonenberg>
so yeah we're fine
<azonenberg>
i was gonna say integralstick doesnt have any GTPs
<azonenberg>
but we have eight LVDS inputs :D
<lain>
also consider cycle limits on those cables/connectors
<lain>
not sure how many cycles SAS cables/connectors are designed to survive
<electronic_eel>
ok, I'm gonna have lunch now and have to work afterwards, so won't be around until it is evening here
<lain>
if the connectors survive many but the cables die after a few hundred, no big deal
<lain>
but if the connectors die, you might have an issue :3
<azonenberg>
lain: hmmmm
<azonenberg>
the molex connector i'm looking at are specified for 250 mating cycles
<azonenberg>
That said, i think we can largely mitigate this if we make the LA and external trigger connector on a passive board that mates with the FPGA board. Which i was planning to do anyway
<azonenberg>
if the connectors wears out this board can be swapped relatively inexpensively
<azonenberg>
also, they're specified for 250 cycles to 10 Gbps
<azonenberg>
i expect for 1 Gsps they'll be good a lot longer
<_whitenotifier-3>
[starshipraider] azonenberg pushed 1 commit to master [+1/-0/±2] https://git.io/Jvbwo
<azonenberg>
we're talking about the logic analyzer input
<Degi>
Okay
<azonenberg>
we're still planning to use usb-c for the active probe power and control
<azonenberg>
but are probably going to use mini-sas (sff-8087) for the LA pod
<Degi>
Hmm okay
<azonenberg>
usb-c doesnt have enough pins
<Degi>
How many diff pairs do we need?
<azonenberg>
Each LA probe will be eight LVDS comparators, an i2c dac for setting Vth, and a tiny mcu for configuring the dac and maybe some sensors
<azonenberg>
so eight diffpairs per bank of 8 LA channels, probably two banks for the whole scope
<Degi>
And the probe will only have the LVDS comparators inside?
<azonenberg>
Correct. no FPGA
<Degi>
Okay
<azonenberg>
just comparators, dac, config mcu, and power conversion
<Degi>
Hm so we'll have to LA frontends and connect them with two SAS cables?
<azonenberg>
That's the plan, yes
<Degi>
Neat
<azonenberg>
each LA bank will be totally independent and able to run at a different voltage
<azonenberg>
both running at 1 Gsps
<Degi>
And a different location
<azonenberg>
Yeah
* Degi
sticks SAS to SATA cable into it
<azonenberg>
i plan to use standard e-z-hook micro grabbers for the LA probes
<azonenberg>
we'll have to figure out some kind of cable to go from those to the LA pod
<Degi>
We could have 2.54 mm pin headers on the LA?
<Degi>
Depending on the distance that could work OK
<azonenberg>
Yeah that will probably work. It's not going to be super fast, probably 125 MHz max give or take a bit
<azonenberg>
i meant more like the physical cables
<Degi>
Those things labelled as "arduino jumper cables" but in fancy maybe
<azonenberg>
the lecroy LA uses some kind of micro coax to the probe ends, so you can put a ground near each channel
<Degi>
Hm
<azonenberg>
I'll probably provide a ground next to each data line and we can figure out details of that lter
<azonenberg>
later*
<Degi>
As long as the loading on the digital circuit isnt too high...
<azonenberg>
Yeah. This will be for lower speed stuff
<azonenberg>
I will also be building a much higher speed LA, 12.5 Gsps
<Degi>
If there was a way to configure a DDR interface on the LA, you could use it to 500 MHz
<azonenberg>
That one will have SMA or MMCX inputs
<azonenberg>
and use the same probes as the analog scope channels
<azonenberg>
probably the transmission line probe i've been working on
<azonenberg>
or a solder in equivalent
<Degi>
Hm maybe a differential probe where there is a balun and two prongs sticking out that you can put on top of a differential trace for capacitive coupling (but the inductance of the balun might be too low for the coupling capacitance, idk)
<azonenberg>
I will be exploring various forms of probe down the road. The transmission line probe is one i understand well and have spent a lot of time optimizing
<azonenberg>
So it will come first
<azonenberg>
The LMH7322, the comparator i plan to use for the LA front end, is $8.20 on digikey @ qty 10, and dual channel
<azonenberg>
So each LA pod in prototype volume will cost me $32.80 plus pcb, stm32f0, dac, and some kind of power conversion to go from 12V down to 5V for the comparator/DAC and 3.3V for the MCU and temp/current/voltage sensors
<azonenberg>
and the enclosure
<azonenberg>
And of course whatever connectors we need
<Degi>
Hmm will each 8 channel unit have a threshold or each channel?
<Degi>
I think it can output 0-5 V on 8 channels at 12 bits
<azonenberg>
interesting. Per channel Vt would be a cool feature
<azonenberg>
i'll look into it later in the day
<Degi>
Like 1 channel is 2.30 € for 1 pc
<azonenberg>
I need to finish the design review for the ADC board and get that ordered first
<Degi>
Oki
<azonenberg>
Then we can start thinking about the LA
<_whitenotifier-3>
[starshipraider] azonenberg pushed 1 commit to master [+0/-1/±4] https://git.io/JvboD
<_whitenotifier-3>
[starshipraider] azonenberg b785d4b - Final hmcad-characterization board as sent to fab
<azonenberg>
you know, ordering these boards now is actually going to work out perfectly on the schedule
<azonenberg>
Because i have some billable work from a customer showing up any day now, i need to assemble their boards and do some bringup work. Having the ADC+AFE boards ordered before that is great
<azonenberg>
because i'll have less time for scopehal work next week