<Famine->
Degi, that's kind of surprising actually, i've nuked a ton of more expensive devices being absent minded and hooking them up with reverse polarity
hlzr has quit [Quit: Connection closed for inactivity]
<lain>
the problem with reverse- and over-voltage conditions on ICs is you don't know what invisible damage may have occurred. I always mark chips as suspect after realizing I've derped up the voltage or etc
<azonenberg>
i have actually seen a few ics, mostly ldos
<azonenberg>
that explicitly state reverse voltage is OK under certain conditions
bvernoux has joined #scopehal
<bvernoux>
hello
<azonenberg>
o/ bvernoux
<azonenberg>
I just ordered a PicoVNA with traceable cal certificate :D
<azonenberg>
Also starting to write an operator manual / datasheet for the probe
<azonenberg>
lain: sooo random idea
<azonenberg>
How do directional couplers work? how wideband can you make them?
<azonenberg>
Would it be possible to make a probe consisting of one half of a directional coupler that's somehow attached to a DUT?
<azonenberg>
using the transmission line being probed as the other half
<azonenberg>
just musing about how to eliminate the unterminated stub you normally get at the tip of a probe between the needle and the first attenuator/buffer stage
<lain>
azonenberg: actually that's how a lot of microwave circuits monitor various paths, they build a directional coupler to siphon off a tiny amount of the signal
<lain>
I don't know how bandwidth calculations work with directional couplers
<lain>
actually it looks like some designs might pass DC
<lain>
but they look like they're narrow bandwidth
<azonenberg>
lain: you misunderstand, i dont mean putting a DC in the circuit at design time
<azonenberg>
i mean something you can press up to a trace and form a DC
<azonenberg>
by coupling to an unmodified DUT
<lain>
yeah, I figured that's what you meant
<lain>
sooo my understanding is this gets a bit tricky due to trace geometry, etc
<lain>
but my gut says you can probably make an /acceptable/ probe that works almost all of the time
<bvernoux>
azonenberg, re ha nice finally you have bought the PicoVNA
<bvernoux>
azonenberg, it seems very nice modern tool
<bvernoux>
azonenberg, do you have bought Cal Kits too ?
<azonenberg>
Yes i got a cal kit
<azonenberg>
and nist traceable certification on both the cal kit and the vna itself
<azonenberg>
gonna be a couple of weeks probably before it shows up
marcos has joined #scopehal
marcos is now known as Guest54665
Guest54665 is now known as marcos__
<bvernoux>
azonenberg, yes nice to do great measurements
<bvernoux>
and it seems ultra fast too
<azonenberg>
Yes. There's a bit of ringing due to ground lead inductance when lain tested with her 5 GHz scope
<bvernoux>
as for example on lot of VNA doing 1601 points for S2P take more than 30s ...
<azonenberg>
it's barely visible on my 2 GHz scope
<azonenberg>
(on my probe
<azonenberg>
)
<azonenberg>
but i wasn't using the low-L ground blade for that test
<azonenberg>
i'm working on drawing up a manual for the probe, designing a template for the calibration certificates, etc
<bvernoux>
do you have comparison of your latest probe with picoProbe or other you have with same setup ?
<azonenberg>
my pico probe came in last night
<azonenberg>
lain has it now and is about to do side by side tests on her 5 GHz scope
<azonenberg>
i haven't tried it on my 2 G yet
<bvernoux>
ha nice a test with 5GHz scope is great
<bvernoux>
it is 20GSPS or 40GSPS ?
<azonenberg>
Only 20 Gsps in realtime mode. It's a LeCroy DDA5005A which is based on the older WavePro i believe. It's a very high end scope but from the windows XP era
<bvernoux>
I speak about the 5Ghz scope how many GSPS ?
<lain>
20 GS/s, but we're testing in RIS mode (equiv. time sampling) at 200 GS/s equivalent
<bvernoux>
ha great
<azonenberg>
My HDO9204 is only 2 GHz but 40 Gsps realtime. Also 200 equivalent
<azonenberg>
We were actually just talking about maybe getting a nist traceable re-cal on her scope so i can do traceable rise time measurements of my probes out to 5 GHz bw
<bvernoux>
also it will be interesting to check the capacitance of the probes
<azonenberg>
although for the moment i plan to market them as 2 GHz probes because i'm confident they perform to spec that afr
<azonenberg>
far*
<bvernoux>
between theory & reality
<azonenberg>
in the next day or two i plan to send out the final revision of my probe PCB to fab
<bvernoux>
it is expected to be something like 5pF max ?
<bvernoux>
or less in fact 2pf ?
<azonenberg>
oh i am expecting hundreds of fF
<bvernoux>
I say when adding everything connector + cable
<azonenberg>
we only care about cap prior to the first resistor, really
<azonenberg>
after that is insignificant loading on the dut
<azonenberg>
my expectation is that the end result will be 500 ohms || 600 fF or thereabouts
<bvernoux>
yes great
<azonenberg>
bvernoux: also, i did some tests of DC resistance
<bvernoux>
and about 2GHz BW ?
<bvernoux>
or more
<azonenberg>
it performs very well out to 2, on a 5 GHz scope there is a bit of ringing with the first ground we tested
<azonenberg>
with a blade/leaf ground you can probably push it further. Also need to re-test on the final pcb revision
<azonenberg>
re DC resistance, my R&S meter is 5 3/4 digit resolution. Cal is a year out of date and i need to refresh it
<azonenberg>
that said, i measured the attenuator on one of my test boards as 449.98 ohms
<azonenberg>
The resistors are only 0.1% tolerance so as low as 449.55 would be within spec
<bvernoux>
but it is still interesting to have results of the R&S meter even if calibration is out of date
<bvernoux>
ha great
<azonenberg>
I intend to ship beta units of the probe with non-traceable DC measurements off this meter, as well as traceable rise time and S2P/bandwidth measurements
<lain>
I forget, does that R&S meter do 4-wire measurement?
<azonenberg>
then get the R&S meter recal'd before shipping production units
<azonenberg>
lain: yes. I did a 2-wire test though
<azonenberg>
i figured at 450 ohms the effects of contact resistance would be pretty insignificant
<lain>
yeah
<azonenberg>
But that was one of the things i was looking for when i bought it
<azonenberg>
since my handheld meter does not
<bvernoux>
if you want to do premium version you can still do selection of best resistors ;)
<azonenberg>
bvernoux: my plan is to ship beta units in essentially production grade status. so final enclosure design, pcb rev, accessories, printed manual, and a full calibration certificate
<azonenberg>
just, as i said, with an out-of-cal DMM
<bvernoux>
azonenberg, ha great
<azonenberg>
but clearly it hasn't drifted much if it shows my resistors as that close to nominal value
<bvernoux>
Yes I doubt out of cal is so far
<azonenberg>
yeah i mean it's R&S, not rigol :p
<bvernoux>
for my VNA calibration showed there was nothing to do
<azonenberg>
i expect nothing but the best
<bvernoux>
even after few years
<azonenberg>
Yeah. i fully expect all of mine are still in spec
<bvernoux>
but for R&S meter it is maybe a bit different especially with such accuracy
<bvernoux>
I should have received my components today by UPS from farnell but no ...
<bvernoux>
again UPS fail ...
<bvernoux>
last time I have wait more than 3 weeks asking 2 times they deliver with nothing and when I have filled a complaint 2days after they have delivered it ...
<bvernoux>
I can understand COVID-19 but why other like DHL, DPD, Chronopost deliver everything in time with at max 1 day off
<azonenberg>
i'm dealing with hardware from a customer that's been taking its sweet time to get through via DHL
<azonenberg>
but that was customs problems, not DHL's fault
<azonenberg>
but also dealing with UPS slowdowns on the probe characterization board from multech
<balrog>
everything is slowed down
<lain>
it's interesting, I bought two things from JP recently. one got here in 2 days via DHL, the other is coming normal Japan Post -> USPS, and it hasn't even left japan :P
<bvernoux>
balrog, my complaint is paying 167Euros for fast delivery which take more than 3 weeks locked at 100km of my house what a shame
<bvernoux>
I'm waiting USPS delivery too
<azonenberg>
side note, my PCBs are coming from multech via UPS
<azonenberg>
Which i've never had happen before
<azonenberg>
they've always shipped DHL before
<bvernoux>
azonenberg, maybe you will be lucky as UPS is failing only in France ;)
<azonenberg>
Lol
<bvernoux>
in my case the issue was just when it arrived in Lyon ...
<azonenberg>
right now it's in Louisville KY
<azonenberg>
which is a bit odd as the package came from china to the EAST coast of the US
<azonenberg>
and now has to go back to the west coast
<lain>
azonenberg: UPS is pretty centralized to their Louisville hub in my experience
<lain>
for int'l shipments
<azonenberg>
i rarely do int'l via ups
<azonenberg>
normally its dhl, occasionally fedex
<Degi>
azonenberg: If you have like 0.2 ohm contact resistance, then it would be 449.78 ohms... I assume you used good banana cables. (My current multimeter has like 0.7 ohms lead resistance lol)
<Degi>
On ZENNECK we should add a 1 GHz BW switch. And I'm not sure but doing the AFE from input to after the splitters single ended would make design a bit simpler I think.
marcos__ has quit [Quit: Konversation terminated!]
marcos has joined #scopehal
marcos is now known as Guest25246
Guest25246 is now known as marcos__
<azonenberg>
Degi: I used the cables R&S shipped
<azonenberg>
the meter allows you to null out cable resistance
<azonenberg>
i forget if i did that for that measurement or not
<Degi>
Hm well
<azonenberg>
if i really want i could do a 4-wire measurement
<azonenberg>
but for a quick one-off i didn't bother
<azonenberg>
for the actual calibration certificates i will probably do 4-wire measurements
<azonenberg>
with a lot of averaging
<azonenberg>
re ZENNECK one of the challenges is that most of the good amplifiers for this sort of stuff are all differential
<azonenberg>
high bandwidth single ended stuff is hard to find
<Degi>
Huh now german mouser is sorted by english again weird
<Degi>
The problem with differential is that we need to split the signal, no idea how to do that differentially
awygle has quit [*.net *.split]
azonenberg has quit [*.net *.split]
azonenberg has joined #scopehal
awygle has joined #scopehal
<awygle>
same way you do single-endedly
<awygle>
just uh
<awygle>
twice
<awygle>
wilkinson power divider
<Degi>
Hm we'd need a resistive divider since DC-wideband design
<Degi>
I mean PCB layout wise
<Degi>
You'd probably need vias or maybe route signal traces below resistors? idk
<azonenberg>
BLONDEL AFE characterization board just got back to oshpark from the fab
<azonenberg>
expecting it to ship out to me today and arrive early next week
<azonenberg>
HMCAD board is a few days behind that
<Degi>
Hm isnt oshpark the fab?
<azonenberg>
No
<azonenberg>
Oshpark is the batching service
<azonenberg>
they take gerbers from dozens of people at a time, tile them onto a panel, and do a single order at a fab
<azonenberg>
you pay per unit area
<azonenberg>
they don't do manufacturing, they're just a middleman
<azonenberg>
But it's a lot cheaper for small boards than going direct to a fab, who is going to want to charge you for a whole panel and all of the setup costs that entiails
<awygle>
don't understand why the pcb layout would be especially difficult
<awygle>
but have not been following
<Degi>
If you want to split a differential signal to two differential signals, how would you do that? I havent found differential splitters yet and if you use single ended splitters, you'd probably need to use vias
<awygle>
yeah a differential splitter is just two single-ended splitters, basically
<awygle>
sec, i sent something to whitequark about this not too long ago...
<Degi>
Yes but how would you route that on a PCB? The splitter outputs would need to cross over somehow...
<azonenberg>
Nothing wrong with vias as long as you match them reasonably well
<Degi>
Hm I mean one side of the pair would need a via while the other wont
<Degi>
Thoough I guess you could put some on the other side for balance
<azonenberg>
Not necessarily
<azonenberg>
I'm thinking, two separate single ended splitters
<Degi>
Yes me too
<azonenberg>
A_P and A_N drop down and cross under
<azonenberg>
B_P and B_N stay on top
<azonenberg>
something like that
<Degi>
Hm the thing from awygle looks neat, but does that work wideband too?
<awygle>
should
<Degi>
Hmm you could do that
<Degi>
Hm what does Zx and Thetax refer to?
<Degi>
You say below that that they're half wave and quarter wave, though our wavelength range of interest is from 30 cm to infinity basically
<Degi>
azonenberg: With your method you'd have some mismatch between different diff pairs but the lines within the pair would be matched... I guess this is not so bad and can be calibrated out with fine gain scaling on the HMCAD.
<azonenberg>
Degi: the pairs would be matched, you'd just have a path length mismatch
<azonenberg>
Which could be easily fixed with routing changes
<Degi>
I mean one pair has vias the other hasnt
<awygle>
can be adapted to resistive pretty easily
<azonenberg>
yes, that's what i'm saying
<azonenberg>
the vias are just a short length of transmission line
<azonenberg>
add 1.6 mm of trace to the other pair and you're good
<azonenberg>
Or just tweak sampling phase on the ADC to calibrate out the propagation delay of the via
<azonenberg>
(which i expect will be needed anyway to cal various things)
<Degi>
Hm well if the vias arent too bad of a mismatch, guess that's fine
<azonenberg>
you forget i can just throw the via geometry in sonnet and impedanec match them
<Degi>
Oh well
<Degi>
Yeah thats useful
<Degi>
As long as inputs and outputs are properly terminated, I guess the only difference will be an amplitude difference at the end
<azonenberg>
not amplitude, phase
<Degi>
Hm rigtjh
<azonenberg>
Which can be corrected just by adjusting the click going to the ADC and/or adjusting trace length
<azonenberg>
clock*
<Degi>
Hm I think all trace lengths should be matched so that even without clock phase alignment it runs pretty good.
<awygle>
still pretty confident you can do this all on one layer lol
<azonenberg>
yes, of course, but there might still be tiny variations
<Degi>
Yes
<Degi>
But that way the variations stay tiny heh
<azonenberg>
awygle: well i'm in no rush. This is about three scopes down the roadmap from where i'm focusing my efforts
<awygle>
sure
<Degi>
Lol... We could connect IN- to OUT1-, IN+ to OUT2+, OUT1+ and OUT2- to GND over 50R and a 200R between IN+ and IN-
<Degi>
That would make half the input swing as common mode voltage though
<azonenberg>
Sounds like a terrible idea
<Degi>
Like have a single mode to differential converter in front of each ADC (we need a diff amp there anyways) and feed one IN- the other IN+
<Degi>
I mean theoretically with ideal components that'd probably work
<azonenberg>
yeah
<azonenberg>
well if you can get me an ideal resistor, i have a bridge i'd like to sell you :p
<Degi>
Is an ideal resistor for infinitesimal delta V enough?
<azonenberg>
Lol i wish
<Degi>
HV resistors are fun..
<Degi>
Now your smudge on it has nonlinear behaviour
<azonenberg>
Degi: my favorite story of crazy component non-idealities has to be that 10M resistor in the harmon instruments board
<azonenberg>
or should i say, the 30 fF capacitor with 10M ohms of leakage resistance? :p
<Degi>
Lol#
<azonenberg>
he got the highest valued resistor he could find
<Degi>
Couldnt he use stripline caps?
<azonenberg>
Not sure
<azonenberg>
for whatever reason, this is what he used
<azonenberg>
i just thought it was hilarious that he used a resistor as a cap
<Degi>
Yes heh
<Degi>
Inb4 using a cap as an inductor
<Degi>
With sufficient voltage you could probably use a resistor as a memristor...
<azonenberg>
With sufficient voltage, all diodes are LEDs ;)
<azonenberg>
(once)
<Degi>
With sufficient voltage, more than once
<azonenberg>
I think at that point it's an arc lamp
<azonenberg>
there isn't any diode left :p
<Degi>
Like a xenon arc lamp but with diode vapors
<azonenberg>
ook so, just to make sure we're on the same page regarding various project status, here's a few updates
<azonenberg>
i'm happy with electrical performance of the probe, expecting to have the final board design for the respin tonight. Still need to spend a bit of time on the enclosure
<azonenberg>
I have a buyer lined up for a decent number of probes right off the bat, details to be announced once it's official
<Degi>
Oh neat
<azonenberg>
So plan is to, in the next day or two, send out the final design to the fab
<azonenberg>
and in the nearish future send out one of my bench DMM's for a traceable cal
<azonenberg>
I'm working on writing a user manual, designing a calibration certificate template, etc
<Degi>
Neat
<azonenberg>
Then i'll probably try and start taking general orders for the probes as soon as the first few are tested, cal'd, and confirmed good
<azonenberg>
use the proceeds from that to fund R&D on BLONDEL
<azonenberg>
all of these characterization boards etc aren't free after all
* Degi
figured out that a single ended design would be possible without any signal splitters
<azonenberg>
I have a PicoVNA ordered and en route
<azonenberg>
next, regarding BLONDEL... I believe all components for the ADC and AFE test board are here
<azonenberg>
as are the solder stencils for both boards, one of the proposed front panel LCDs, and one of the Hammond enclosures
<azonenberg>
which i've opened the box for but not screwed together
<azonenberg>
AFE test board is ETA early next week, ADC board shortly after that
<azonenberg>
It comes with a default manufacturer cal but at the time of purchase for an extra $300 you can get a full traceable cal with all of the data
<azonenberg>
Or you can take it to a cal lab yourself and have that done instead
<Degi>
And how would large scale assembly go? Like if a lot of scopes need to be produced, how do you: Solder the PCBs, assemble the cases, do testing
<Degi>
Neat
<azonenberg>
Not sure on scopes yet. Probes i am going to hand assemble at least for the short term, unless i move a lot of volume
<azonenberg>
six resistors, three connectors are easy enough to do with tweezers
<azonenberg>
i can probably populate 20+ probes and run them through the oven in a batch
<azonenberg>
then cal the whole lot
<Degi>
Hm yeah
<azonenberg>
Re BLONDEL floorplan green board is the main board, which basically needs to be that size/shape in order to have rear-panel connectors and also overlap the front-side boards so connectors mate
<azonenberg>
if it's much wider than shown, it won't fit in my current oven
<azonenberg>
the pink boards all mate with it
<Degi>
And another suggestion: Moving azonenberg/starshipraider to starshipraider/<subprojects>
<azonenberg>
rightmost is the socket for the trigger/LA module
<azonenberg>
middle two are ADC+AFE boards
<azonenberg>
leftmost is either a rigid, flex pcb, ribbon cable, or similar from the lcd to the main board
<azonenberg>
mostly passive, may contain a backlight boost converter
<azonenberg>
Main board will contain the artix-7 fpga, the stm32 to run the display and scpi stack plus control all of the sensors, i2c/spi buses, etc, a sodimm of ddr3 for capture buffer, an ethernet phy
<azonenberg>
and as an optional addon, a tlk10232 and 10G SFP+ cage
<azonenberg>
and the PLL
<azonenberg>
as well as power supplies for all of the digital logic and possibly the comparator/dac for the external trigger, not sure if that will be on the trigger board or main board yet
<azonenberg>
Once all of the current boards come in, i plan to slow down on hardware work for a bit. I have firmware to write for the stm32 on BLONDEL, a scopehal driver for the prototype
<azonenberg>
we need to define the scpi command set for the control plane, figure out a protocol for the data plane (it can't be just raw adc samples over tcp, there needs to be some header for trigger time, channel number, number of points, etc)
<azonenberg>
And this will give my wallet a chance to recover from all of the hardware i've just ordered :p
<Degi>
Maybe something where the first byte indicates packet type and the rest is ADC data or control data?
<azonenberg>
My thought is to have a framing header for each trigger event
<azonenberg>
timestamp, list of channels with data
<azonenberg>
for each channel sample interval, number of points, then raw data
<Degi>
Maybe the ADC settings too?
<Degi>
Hm damn I should've put a USB UART interface on the PCIe port. That way I wont have to use the dongle I have now and could directly connect a cable to it
<azonenberg>
usb pcie dongle? :p
<Degi>
Haha no, the ECP5 EVN to PCIe adapter
<azonenberg>
This reminds me of a bin we have on a shelf at $dayjob
<azonenberg>
there's a whole shelf full of USB cables of various sorts
<azonenberg>
A-B, A-mini B, A-micro B, A-C, etc
<azonenberg>
And one bin labeled "USB Abominations"
<azonenberg>
which consists of pretty much every perversion of USB you can think of that either came with some gizmo or had to be cobbled together to talk to a customer's product
<Degi>
Uggggh is firefox really using 75% of RAM... That is like 12 GB
<azonenberg>
usb a-a male-male
<Degi>
Lol I have a A-A too
<azonenberg>
usb to rj45 (not ethernet, literally 8p8c crimped onto a usb cable)
<Degi>
Not sure if its passive or active tho
<Degi>
Hahah
<azonenberg>
one of my favorites is a 3-terminal cable
<Degi>
USB to 3 phase power outlet too?
<azonenberg>
one end is a usb a male
<azonenberg>
the other side has usb a and mini-B male
<azonenberg>
i haven't got a clue what it was for
<Degi>
Hm maybe some kinda thing to provide power for HDDs or so?
<azonenberg>
i've seen ones that had two A's at one side to steal extra power for something
<Degi>
Usually they'd have both As on one side tho
<azonenberg>
exactly
<azonenberg>
this is A -> {A, mini-B}
<Degi>
Maybe it came with a USB wall wart
<azonenberg>
but why the second A?
<azonenberg>
it just makes no sense
<Degi>
One A to the wall wart, the other to the PC
<azonenberg>
we have {A,A} -> microB
<azonenberg>
i've see nthat before
<azonenberg>
but one A at each end PLUS another cable?
<Degi>
So you can have the PC and HDD nearby and the wall wart somewhere else
<azonenberg>
then there's all manner of proprietary connector
<azonenberg>
probably mostly predating mini/micro
<Degi>
Or wall wart and HDD nearby and PC further away
<Degi>
I have a GPIB cable but the device I wanted to use it on doesnt boot anymore
<Degi>
And some weird fiber optic cable with something that looks like a mini version of BNC connector?
<monochroma>
ST or FC connector
<Degi>
I think its ST
<monochroma>
yeah, that used to be extremely common
<Degi>
Its SMA to ST x2 then I guess
<Degi>
Hmm I do kinda wonder how many watts fiber-optic cables can take...
<miek>
i was trying to use a usb to rs232 adapter the other day, till i eventually found it was just usb wired to d-sub :(
<Degi>
Lol
<Degi>
Why tho...
<Degi>
Is there D SUB with fiberoptic wires too?
<Degi>
There is
<monochroma>
what's even more confusing is fiberchannel used to have a copper based dsub 9 interface version :P
<Degi>
Why is D-SUB even called subminiature? Like the smaller version of what?
<Degi>
Oh they were the smallest at their time... I do kinda wonder if people back then just didn't have the manufacturing ability for stuff similar to micro USB
<monochroma>
correct
<Degi>
Hm there's D sub for RF, fiber optic, power... Is there D sub for cooling water too?
<monochroma>
Degi: DB13W3
<Degi>
I mean thats just coax tho?
<monochroma>
ah no i wasn't responding to water cooling lol
<Degi>
I want something like DB13H2O2
<monochroma>
was just pointing out an interesting connector that used to be extremely common
<Degi>
Heh DB13W3 is basically VGA with some differences, right
<monochroma>
sorta
<monochroma>
sometimes semi-compatible
<monochroma>
but yeah, basicaly high quality RGB + a control channel
<Degi>
Maybe I could get one and drill out the coax plastic and add some rubber rings lol
<Degi>
Huh there is liquid cooling for servers... But like with fans within the blade. Kinda seems like a bad idea to me...