<Famine>
o.O this is seriously a hammer in search of a nail. reading a paper on low noise preamps, the design calls for 10 ADA4898-2 op amps in parallel. that's only ~$100 in op amps
<azonenberg>
lol
<Famine>
azonenberg, although he is down to 200 pV/root Hz
<azonenberg>
what is he using it for?
<azonenberg>
Most of my research on frontends etc has been going for bandwidth, not super weak signal performance
<azonenberg>
i think i have basically all of the metadata for (supported subsets of) scope and channel config, UI layout, etc. What's missing is the descriptors for the WaveformArea objects (right now i define the tree hierarchy of splitters, but not what goes inside each pane)
<azonenberg>
as well as waveform data itself
<azonenberg>
may make a few minor tweaks but how's it look?
<azonenberg>
(ping: lain)
* lain
looks
<lain>
azonenberg: uhhhh I'm sleepy but it looks quite good!
<awygle>
request ability to save just the waveform data as well
<azonenberg>
awygle: that's already there
<azonenberg>
well, the arhcitecture is
<monochroma>
i think that was the plan, you can save the layout + waveform or just the waveform
<azonenberg>
well, ok
<azonenberg>
You can save layout only, or layout + waveform
<azonenberg>
When *loading* you can choose to import layout, waveforms, or both
<azonenberg>
but the file will always contain layout
<awygle>
right, i may not want layout on occasion
<azonenberg>
Yes that's a load-time decision
<azonenberg>
But the layout is there in the file if you want it
<awygle>
i'm envisioning exporting to some other program
<azonenberg>
So, there will also be a mode to export raw waveforms as CSV or similar
<awygle>
okay, that's all i really want
<azonenberg>
The overall hierarchy will look like this
<azonenberg>
foo.scopesession: metadata for the instrument setup and UI config, protocol decoders, etc
<azonenberg>
foo_data/$instrument/waveforms.yml: metadata for each waveform (probably just timestamp)
<azonenberg>
foo_data/$instrument/waveform%d/$channel.wave: raw binary waveform data
<azonenberg>
oh and waveforms.yml will also include sample rate metadata for each waveform in case you reconfigured the scope halfway through a session
<azonenberg>
offset and duration values are in sample-clock units
<_whitenotifier-3>
[scopehal-docs] azonenberg opened issue #4: Document save file format - https://git.io/JvV8h
<_whitenotifier-3>
[scopehal-apps] azonenberg pushed 1 commit to master [+0/-0/±2] https://git.io/JvV4m
<_whitenotifier-3>
[scopehal-apps] azonenberg b33d998 - WaveformAreas are now mostly serialized. Still need to do measurements but UI config is otherwise done.
<_whitenotifier-3>
[scopehal-apps] azonenberg pushed 1 commit to master [+0/-0/±3] https://git.io/JvV4B
<_whitenotifier-3>
[scopehal-apps] azonenberg 202ff55 - Refactored configuration serialization for WaveformGroup out into a separate function. Began work on measurement serialization. See #3.
<_whitenotifier-3>
[scopehal] azonenberg labeled issue #83: Allow measurements to return one value per UI instead of just one for the whole waveform - https://git.io/JeQQs
<_whitenotifier-3>
[scopehal] azonenberg commented on issue #83: Allow measurements to return one value per UI instead of just one for the whole waveform - https://git.io/JvV4V
<_whitenotifier-3>
[scopehal] azonenberg pushed 1 commit to master [+0/-0/±2] https://git.io/JvV45
<_whitenotifier-3>
[scopehal] azonenberg 419dc0f - Measurement: initial serialization support
<_whitenotifier-3>
[scopehal-apps] azonenberg pushed 1 commit to master [+0/-0/±1] https://git.io/JvV4F
<_whitenotifier-3>
[scopehal-apps] azonenberg a0bab26 - WaveformGroup: now serialize measurements (see #3)
<azonenberg>
ok i think this is just about all of the config left
bluezinc has quit [*.net *.split]
apo has quit [*.net *.split]
gruetzkopf has quit [*.net *.split]
wbraun has quit [*.net *.split]
awygle has quit [*.net *.split]
m4ssi has quit [*.net *.split]
maartenBE has quit [*.net *.split]
lain has quit [*.net *.split]
Ekho has quit [*.net *.split]
sorear has quit [*.net *.split]
elms has quit [*.net *.split]
_whitenotifier-3 has quit [*.net *.split]
_whitelogger has joined #scopehal
LeoBodnar has joined #scopehal
futarisIRCcloud has joined #scopehal
Famine has joined #scopehal
<azonenberg>
Getting 50 ohm traces on 1.6mm RO4350B is turning out to be a bit of a chore. I'm now using 0.8mm trace with 0.1mm gap as a GCPW
<azonenberg>
and the SMA match is being rather fun too
<azonenberg>
Seems i get the best results with 0.15mm wide gap inside the SMA footprint. So kinda like one of those fancy Rosenberger connectors bvernoux is a fan of
<azonenberg>
while it's technically a GCPW, the ground is of almost no importance at this substrate thickness
<azonenberg>
Substantially all of the return current flows in the top layer
<azonenberg>
I get better than -30 dB return loss from 100 MHz - 10 GHz so that is probably good enough for the characterization board
<azonenberg>
optimizing impedance further seems futile as the fab can't control to +/- 0.1 ohm anyway :p
<azonenberg>
I should probably sleep soon, but tomorrow after work i hope to get the characterization board banged out. It should be quick... eight SMAs and four resistors plus a probe connector
<Famine>
i have no idea why i'm up this late either, drooling over some rosenberger SMAs
<azonenberg>
Lol
<azonenberg>
because i mentioned them, or independently?
<azonenberg>
if you just happened to be drooling over them without me mentioning them that would be impressive
<azonenberg>
and a sign that you're as huge of a nerd as me :p
<azonenberg>
(this footprint is designed for SMA-J-P-H-ST-EM1 which is a whopping $3.80 @ qty 25)
<Famine>
azonenberg, independently, 32K243-40ML5 and 32K242-40ML5
<azonenberg>
The standard samtec footprint is clearly designed for microstrip with a ground plane << 1.6mm from the signal layer
<azonenberg>
but i didn't feel like paying for *multilayer* rogers
<Famine>
the 32K242-40ML5 is 4.19 @ 10 pieces
<azonenberg>
So i decided to engineer a match that would work on 1.6mm RO4350B
<Famine>
i'm pretty comfortable <200MHz, you really can't screw up when your this slow lol
<azonenberg>
yeah at that frequency you can send your signal over barbed wire and it wont care :p
<Famine>
who do you think i am ? i only use the highest quality chicken wire!
<azonenberg>
lol
<Famine>
so you were saying you had a bunch of issues with over etching ?
<azonenberg>
When? on what process?
<azonenberg>
i had overetch issues on oshpark in 2011, yes :p
<Famine>
i'm a little behind the times then :P
<azonenberg>
Or you were thinking of somebody else
<azonenberg>
but no, what i'm trying to do here is make the characterization board for fine tuning the transmission line probe
<azonenberg>
The probe uses a mixed RO4350B + cheap FR4 stackup because I wanted to keep the transmission line at 0.5mm width for matching to the resistors
<azonenberg>
Perhaps not strictly necessary but it did keep layout simple
<azonenberg>
with 0.1mm ground i needed a fairly thin prepreg
<Famine>
quite possibly, i've been reading a ton of papers/blogs on RF design and low noise measurements
<azonenberg>
0.1mm spacing to ground*
<azonenberg>
then a cutout under that for the large, fat sma center conductor
<azonenberg>
on the characterization board i have a very different problem, the ground plane is so far away as to be practically nonexistent
<azonenberg>
almost all return current flows in-plane
<azonenberg>
And i had to enlarge the signal conductor to 0.8mm in order to get a reasonable impedance without having the ground be super close, 0.1mm is already tight
<azonenberg>
then i just had to tune for the center conductor, which needed 0.15mm space
<Famine>
i need to keep reading more books until its not black magic voodoo
<azonenberg>
i've found running lots of simulations helps to get a more intuitive feel for how things work as well
<azonenberg>
not in terms of being able to get exact numbers or anything, but in terms of being able to predict cause and effect of various changes to layout
<azonenberg>
even the lite version of sonnet is likely going to be helpful for that
<azonenberg>
(btw did you register it? the ram cap is 1 MB unregistered and 32 registered... yes, they get your email address, but they don't really spam you)
<Famine>
yea, i had to manually email them the automated registration was down
<Famine>
interesting NIST still hasn't changed its recommendation on RSA key sizes, but these are the same guys that argued for smaller symmetric keys for years
m4ssi has quit [Remote host closed the connection]
<awygle>
yeah at 1.6mm thickness the backside is so far away as to be irrelevant for TL purposes
<awygle>
also you shoulda used SMA-EM3 :p
<awygle>
and SMA-EM1 is $2.48 @ qty1 direct from Samtec. always buy samtec direct, digikey et al appear to be re-selling at a markup
<azonenberg>
that is their entire business model :p
<azonenberg>
when you're not buying a lot, saving shipping is worth it. and will samtec sell you less than like case quantity direct?
<azonenberg>
most part vendors only sell by the reel etc
bvernoux has joined #scopehal
<awygle>
yes but usually they're authorized distributors not just dudes that buy from the Samtec website, which seems to be hte case here
<awygle>
yeah they'll sell you whatever qty. i order 5 SMAs from them all the time (or did in a former life)
<awygle>
and i don't think they ever charged me shipping, and it was there in 2 days
<azonenberg>
oh cool
<awygle>
one of many reasons i always recommend them
<awygle>
but anyway yes, SMA-EM3 is much cooler :p
<bvernoux>
it is translated from chinese to english as I'm not fluent in Mandarin ;)
<azonenberg>
welp, in case there was any doubt i've been bitten by the RF bug
<azonenberg>
i just ordered a rosenberger sma-bnc adapter
<azonenberg>
$20 for one adapter :o
<awygle>
lol
<azonenberg>
But i'm having an issue that i think might in part be due to mismatch of my current sma-bnc adapters
<azonenberg>
And i want to eliminate variables to narrow down the source of it
<miek>
the sonnet license was a bit of a giveaway too :p
<azonenberg>
Lol
<azonenberg>
the bigger giveaway isn't that i bought sonnet, it's that i want more after buying the basic edition :p
<miek>
hah
<awygle>
time to design a PIFA
<azonenberg>
a what?
<azonenberg>
power isolated foobar array? :p
<monochroma>
a pain in the fscking a$$?
<miek>
planar inverted f antenna
<awygle>
^
<monochroma>
what for? :o
<awygle>
ionno. flex?
<awygle>
just seemed like a good use of sonnet
<monochroma>
hehe
<azonenberg>
i dont actually have the far field radiation feature
<azonenberg>
So i couldn't if it wanted to
<azonenberg>
Although they'd gladly sell me the option if i asked
<awygle>
design me a coplanar waveguide for PCIe on 2L FR4
<awygle>
i long to use differential CPW on a cheapo 2L board someday.... just need to make time for one of the projects that could use that
<azonenberg>
awygle: give me detailed info on stackup etc (is there a ground plane? how far out? soldermask? what Er? what copper thickness?)
<azonenberg>
do you just want 100 differential or does it have to be 50 single ended too?
<azonenberg>
min trace/space i can use?
<awygle>
100 ohm differential (SE not important), 1.6mm thickness, 4/4 min (6/6 would be better but :shrug:), soldermask, Er 4.6, ground plane on opposite side, 1oz copper on both sides
<awygle>
+/- 10%
<awygle>
huh actually 6/6 with 2oz copper is pretty close to dead-on
<azonenberg>
Does that model thick metal?
<awygle>
yeah'
<awygle>
i'm winging the roughness at 0.1 um but the rest of hte params seem to be reasonable. this is at 1 GHz which is probably the wrong frequency tho
<awygle>
doesn't matter too much
<awygle>
i would be interested to see if a sim of this came out comparable
<azonenberg>
i have a sweep running with a few different parameters, give it a bit
<awygle>
with 4/4 space and 1oz copper i get W=5mil as a pretty good answer
<awygle>
or 6 space, W=10 on 1oz, which is also fairly reasonable if a bit wide
<azonenberg>
are you doing outer ground plane too? or just edge coupled
<azonenberg>
CPW implies a ground on the outer layer, otherwise it's differential microstrip
<electronic_eel>
I would just buy racks with square cage nuts now. Because some vendors have rackmount kits which don't screw into the cage nuts, but have some clever snap-in mechanisms that replace the cage nuts
<azonenberg>
awygle: ah nice
<electronic_eel>
and those usually just work with the square ones
<awygle>
true. fortunately none of my present equipment has that problem though.
* azonenberg
was just drooling over the R&S NGP800 series. I have no rack space for it and absolutely can't justify the money
<azonenberg>
but it's a sexy piece of hardware
<awygle>
and i'm at the limit of the single-rack lifestyle
<awygle>
so it'll be fine, if i get something new it'll just go in the 8U or 12U, or a new rack
<azonenberg>
yeah i have two racks now
<azonenberg>
left rack has the HDO, right has the waverunner, then a HMC8012 + HMC8042 side by side below the scope
<azonenberg>
it's a bit bulkier than it strictly needs to be due to cable management panels but they're absolutely worth it
<azonenberg>
when you have scope probes NOT dangling in front of your psu :p
<awygle>
SA: 6U. Scope: 4U. PSU: 4U.
<azonenberg>
4u psu? o_O
<azonenberg>
well that's your problem
<azonenberg>
also 4U scope is quite short, even my old rigol was 5 or 6 when racked
<azonenberg>
although it was admittedly not a very dense rackmount
<awygle>
yeah all the rigol/siglent rack mounts are really just like... shelves with some fiddly bits
<azonenberg>
meanwhile my psu is half width 2U, lol
<azonenberg>
I guess thats what you get for getting a switcher with a linear postregulator instead of a pure linear
<awygle>
we had these half-rack-half-u sorensons at planetary, they were great
<azonenberg>
yeah i have 2ch psu + 5.75 digit dmm side by side in 2U
<electronic_eel>
my PSUs are 1/3 of 19" and 3U, so they line up nicely left of my scope (on the same shelf, not racked)
<azonenberg>
My scopes are something like 8-9U *each*
<azonenberg>
which is one of the reasons why i want to build a proper headless DSO
<azonenberg>
that's a LOT of bench space to be wasting
<awygle>
they were also literally an order of magnitude more expensive than mine lol
<awygle>
if not two
<azonenberg>
Unless you have a $150 scope, probably not two
<azonenberg>
maybe 1.5
<awygle>
350$. not quite.
<azonenberg>
"your scope is ~17 dB cheaper than mine"?
<awygle>
ha
<azonenberg>
But no that was not me bragging about having nice hardware, that was me complaining about how big it is
<azonenberg>
If i could get the same scope electronics in a 2U headless i would
<electronic_eel>
yeah 9U is enormous. replacing that with regular monitors, maybe on some holding arms, would improve that a lot
<awygle>
course
<azonenberg>
in fact, lecroy *makes* the waverunner 8000-R series but it's the same exact scope sans monitor
<azonenberg>
and i couldn't find one secondhand for the same price i could the full unit
<azonenberg>
electronic_eel: this thing literally has a 720p 15" touchscreen on it
<electronic_eel>
hmm, but would you want a touchscreen in front of you? wouldn't some monitor on an holding arm above you, and mouse+keyboard, be better?
<azonenberg>
10" display instead of 15" that the new one has and i think 1 or 2U shorter
<azonenberg>
that's the whole point of glscopeclient, lol
<azonenberg>
i dont like touchscreen scopes
<azonenberg>
i want to replace these scopes with a headless 1U/2U that's just AFEs, ADCs, and a 10G or 40G ethernet controller
<azonenberg>
and free up all that rack space for more gear
<electronic_eel>
aren't the lecroy scopes effectively windows pcs with a scope app and a internal interface to the scope hw?
<azonenberg>
Yes
<azonenberg>
typically it's pcie via a custom riser card etc
<azonenberg>
these run win 7 embedded pro, newer are win10
<azonenberg>
both of my current scopes have dual core i5s, one is 8GB DDR3 the other is 16
<azonenberg>
which is an awful lot of fluff, because they're running the whole scope UI etc there too
<azonenberg>
i like the unix philosophy of test equipment: a scope should be an ADC going to a pipe
<azonenberg>
:p
<azonenberg>
you should not have UIs on the hardware, you should have APIs
<miek>
i think my scope has a PowerPC 405
<azonenberg>
And when i build one it won't even have a cortex-m0 :p
<awygle>
i thought $work was gonna pay you to work on starshipraider
<azonenberg>
awygle: there was discussion of it. But billable stuff takes priority
<azonenberg>
i want to get back to SSR soon and try to plug it into scopehal as a LA
<awygle>
sure
<azonenberg>
among other things
<awygle>
what happened to latentred and integralstick? did those cross the finish line?
<azonenberg>
now that i have a higher b/w scope and such i should have a much easier time characterizing the i/o boards
<azonenberg>
integralstick is substantially done, i'm using it on several projects now
<azonenberg>
The pcb needs a tweak at some point to fix the bad mac address eeprom footprint but i have a big stock of pcbs i'm just reworking with deadbugged eeproms on an as needed basis
<azonenberg>
so i'm in no hurry to respin
<awygle>
and that ended up being a spartan 7 and a cortex m3 or something?
<azonenberg>
artix7 + cortex-m7
<awygle>
ah
<azonenberg>
ftg256 artix7 your choice of density, stm32f777 (not sure what else might be pin compatible, i've only used that), dual channel 1.8v hyperram, spi flash, rgmii phy, dc-dc to run on a single 5V supply
<azonenberg>
and a wide range of 3.3, 2.5, 1.8, and lvds io
<azonenberg>
latentred will use integralstick as the management engine
<azonenberg>
i've written a decent bit of the CLI already
<azonenberg>
a fair fraction of the switch fabric IP was developed and tested on a virtex devkit, i have yet to spin the PCB with the final FPGA on it but hope to do that this month into next
<azonenberg>
the line card pcb is ~90% complete but needs to have the backplane connector swapped around
<azonenberg>
i'm switching from a star to daisy chain topology to simplify the pcbs and not need a 19" wide PCB
<azonenberg>
there will be a passive jumper board kinda like a SLI bridge that straddles the switch engine to the first line card, first to second, and second to third
<azonenberg>
each line card will have 8 sgmii phys and have connections for 24 lanes of sgmii
<azonenberg>
first 8 go on chip, the other 16 get pushed to the line card at left
<awygle>
are you redriving? sounds like you could have SI problems if not
<azonenberg>
I have two different redrivers in mind, one is just a buffer and the other does full CDR+retiming with equalization and preemphasis
<azonenberg>
my plan is to do the switch engine board first then make some passive boards with long backplane traces to measure actual SI in system through a bunch of connectors etc
<azonenberg>
then make the final decision as to what buffering, if any, the line cards will have for the passthrough sgmii lanes
<azonenberg>
that was the other holdup, i was originally going to wait until freesample was ready
<azonenberg>
but my new lecroy has the bandwidth to do sgmii SI work
<azonenberg>
i routinely reshuffle projects' relative priorities as the dependency graph changes
<azonenberg>
the nearest term priority now is the probe because i will be using it to take SI measurements on the backplane etc
<azonenberg>
next priority once i've got that under control is the switch engine board because i need something to do dev on for the rest of latentred, plus some openfpga folks are interested in using it as a testbed for symbiflow on 7 series
<azonenberg>
it will also give me a data source for lots of fun stuff to write scopehal decoders for, like sgmii and 10gbase-r
<azonenberg>
as well as lots of random i2c sensors etc i can use to test some of that side of the framework
<azonenberg>
also a good testbed for multiple phys on one mdio bus
<azonenberg>
Which is something i've needed to improve support for for a while in scopehal, i just lacked any data to test on