<azonenberg>
lain: so i reproduced your weird eye issue
<lain>
:o
<lain>
is it a bit depth issue?
<azonenberg>
I didnt say i fixed it
<azonenberg>
i reproduced it :p
<azonenberg>
i'm just starting to look into it, have to finish massaging some pending code changes and commit them first
<azonenberg>
I'm working on like five things at once
<lain>
:D
_whitenotifier-c has joined #scopehal
<_whitenotifier-c>
[scopehal-apps] azonenberg pushed 3 commits to master [+0/-0/±6] https://git.io/JflhW
<_whitenotifier-c>
[scopehal-apps] azonenberg 6ccc407 - WaveformArea: added more space for right side axis labels
<_whitenotifier-c>
[scopehal-apps] azonenberg 6528cf5 - Stats frames are now properly hidden in new waveform groups without stats
<_whitenotifier-c>
[scopehal-apps] azonenberg 83e8a0e - Protocol decoder dialog no longer blocks background waveform processing and rendering while displayed. Fixes #91.
<_whitenotifier-c>
[scopehal-apps] azonenberg closed issue #91: Protocol decoder dialog should not freeze entire UI when active - https://git.io/Jfl2h
<azonenberg>
also idk if you've played with maui studio yet but it's quite fun. you can import "labnotebook" save files from real scopes then analyze them with a pretty good complement of software options on any of eight simulated scopes up to and including a labmaster
<azonenberg>
interestingly the pam4 decode option is not included in the demo bundle
<azonenberg>
pam4 analysis*
<azonenberg>
its the only one i've found that isn't there
<azonenberg>
SDA-II doesn't seem to support multiple eyes at all
<azonenberg>
SDA-III does, but the UI is REALLY clunky
<azonenberg>
its hard to describe, but i dont like how they did it at *all*
<azonenberg>
you can use SDA to display eyes on MLT3 if you set the trigger levels right but all of the analytics assume NRZ
<azonenberg>
So it can't measure opening etc
<lain>
oh neat
<azonenberg>
if you use the "automotive ethernet" option you can do full eye analysis but the eyes look different because it's not using the same analysis path as SDA
<azonenberg>
and this options does not support multiple eyes
<azonenberg>
as far as i can tell even with the simulated labmaster, it is not possible to simultaneously display two MLT3 eyes, e.g. tx and rx, and plot opening measurements on both
<azonenberg>
also as far as i can tell it only lets you display multiple SDA eyes in the "q-scape" tabbed display mode
<azonenberg>
which basically means you either have to tab between different lanes, or display two or more identical scope grids side by side
<azonenberg>
there does not seem to be *any* way to have such a simple structure as two eyes side by side with waveforms and decodes around them
<azonenberg>
oh and with sda + qscape, you have to have the eyes in the same region of the display
<azonenberg>
you cant have one in the top right quad and one in the bottom right quad on the two views
<azonenberg>
i'll grab some screenshots at some point of side by side comparisons
<azonenberg>
but basically as long as you only use it in a few narrow configs, maui looks really nice
<azonenberg>
but it's super constraining and not at all flexible if you want to adjust the UI to do anything but their canned workflows
<azonenberg>
lain: so here's how the eye processing works now (scopeprotocols/EyeDecoder2.cpp:398 and so on
<azonenberg>
Loop over all samples in the incoming waveform
<azonenberg>
for each sample, find the offset to the clock (nominally centered in the UI)
<azonenberg>
x position of each pixel is that offset converted to a pixel value
<azonenberg>
y position is the actual voltage of the sample plotted just like in a normal scope view
<azonenberg>
I think what's happening is that there's beats between the resolution of the eye and the resolution of the waveform, in both time and voltage
<azonenberg>
i actually have a hack in EyeDecoder2.cpp:111 that helps
<azonenberg>
i think the failure mode is if we skip two rows of pixels
<azonenberg>
this is what the eye looks like without the hack
<azonenberg>
maybe there's a way to interpolate to avoid this being needed
<azonenberg>
here you can see we seem to have a pattern of 2 lines valid, 1 blank
<azonenberg>
with this particular combination of eye size and adc range
<lain>
just call it scanline simulation and ship it
<azonenberg>
lol
<azonenberg>
So what i think i might do is antialias in both x and y
<azonenberg>
rather than rounding and incrementing one single pixel i'll calculate coverage and spread the bump over 2x2 pixels
<azonenberg>
the eye is int64 based to avoid numerical integration problems, but i think i can solve that by bumping the eye by (say) 10 instead of 1 counts per sample
<azonenberg>
which will give me some resolution for AA
<azonenberg>
To start, since the PLL provides some level of AA in the x axis, i'll only AA in the Y axis
<azonenberg>
i think the AA is the right idea but some kind of interpolation is still needed because i still have 256 discrete Y values, they're just no longer at integer Y coordinates
<azonenberg>
i guess one "cheap" option is to do a denoising filter on the eye after integration before displaying it
<azonenberg>
but i dont want to obscure actual details with a median filter etc
<_whitenotifier-c>
[scopehal-apps] azonenberg opened issue #93: Better handling of file loads when "load waveform data" is selected, but there's no waveform data in the save file - https://git.io/Jfljg
<_whitenotifier-c>
[scopehal-apps] azonenberg labeled issue #93: Better handling of file loads when "load waveform data" is selected, but there's no waveform data in the save file - https://git.io/Jfljg
<azonenberg>
and then we still have the weird vertical line down the middle at the zero point
<_whitenotifier-c>
[scopehal-apps] azonenberg opened issue #94: Hide statistics when loading a save file that doesn't have them enabled - https://git.io/JfljP
<_whitenotifier-c>
[scopehal-apps] azonenberg labeled issue #94: Hide statistics when loading a save file that doesn't have them enabled - https://git.io/JfljP
<Degi>
Does the usb isolator even have analog capability?
<Degi>
And how to supply power, I'll need to find / build a transformer with 5 kV insulation heh
<Degi>
Oh Digikey has good transformer search
<Degi>
Well, good enough.
<Degi>
Meh they need very low frequencies of like 100 KHz, maybe a planar transformer will work
<azonenberg>
Welp
<azonenberg>
probe dummies are out for delivery just in time for the actual probes to be on the way. Using UPS, we'll see if they are faster than fedex :p
<azonenberg>
i already got a call about customs clearance which suggests they're moving along faster
<Degi>
Yay!
<azonenberg>
still have a bunch more work to do on the DDR RAM side of things
<azonenberg>
but i have a long list of older relatively-easy tickets i'm trying to close out first
<azonenberg>
probably ethernet autonegotiation, that seems like a pretty simple one
<azonenberg>
i have a decode for it already but it's bitrotted
<azonenberg>
so i have to do whatever refactoring is needed to make it work with current scopehal
<azonenberg>
probably going to do one for SPI soon too because, to be honest, it's long overdue
<azonenberg>
shows our priorities, having 8b10b and ddr ram decodes before spi :p
<Degi>
Ah yes, "no clean" flux...
* Degi
proceeds to clean off the flux
<Degi>
How bad is the nub on vertically mounted pcb SMA conenctors?
_whitelogger has joined #scopehal
<azonenberg>
you mean the unterminated stub on PTH SMAs?
<azonenberg>
how far out does it stick past the layer your signal lives on?
<Degi>
Like 4 mm
<Degi>
The worst case is I have trace on layer 1 and SMA is directly on top of it
<Degi>
I think ill dremel off the stubs
<Degi>
And I forgot to remove the copper fill from around the center pin, I hope that isnt too bad
<azonenberg>
So a 4mm unterminated stub
<bvernoux>
azonenberg, I have seen announcement of a new PicoVNA 8GHz
<bvernoux>
azonenberg, maybe it is good time to ask them to provide a decent PC APP ;)
<Degi>
Huh the ground plane capacitance actually makes my very bad ltspice model look better. With 100 fF I had like 3.3 dB loss with 800 fF I have like 1 dB loss
<azonenberg>
Degi: that's going to be a quarter wave stub around 15 GHz if my math is right
<Degi>
Frequency of concern: 2.5 GHz
<azonenberg>
so you'll get a null there, and rolloff towards it
<azonenberg>
at 2.5 GHz it probably wont have too big an impact
<azonenberg>
lambda there is 30 mm
<Degi>
Ok ill still dremel it off I think
<Degi>
You mean quarter lambda?
<azonenberg>
yes
<azonenberg>
no
<bvernoux>
ha no it is PicoSource ;)
<azonenberg>
sorry following two threads at once here
<bvernoux>
I was confused by the photo with same case as PicoVNA ;)
<Degi>
Kinda feel like my PCIe signal would look very rounded, if I had a good scope.
<bvernoux>
the idea is to compare with same setup VNA, calibration different hardware which do similar things a bit like here with GCPW vs Attenuator
<bvernoux>
SO anyone interested in improving Qucs stuff all is provided ;)
<bvernoux>
Idea is to find good paramter for Qucs simulation to have something which is not too far from reality with different PCB
<bvernoux>
The idea behind is to add more and more stuff like complex microwave filters to check how to optimize them ;)
<bvernoux>
and to test them as far as possible up to 6GHz so far with my VNA but later(when I will have repaired my VNA and also received my KC908) I can also test up to 26.5GHz with my SA ;)
<miek>
bvernoux: i got my JFW attenuators by the way. looks like they were never used, they seem to be new old stock that had sat on an MOD shelf for a while :)
<bvernoux>
miek, ha nice I'm waiting mine too I shall receive 2 ;)
<bvernoux>
miek, what is the ref for yours ?
<Degi>
MOD shelf?
<bvernoux>
JFW do lot of custom version in fact
<bvernoux>
I will receive one which is not standard too
<bvernoux>
I shall receive "JFW Industries Inc RF attenuator Model 50BR-001. 0 to 110db 50 Ohm 2GHz"
<miek>
50DR-045 - couldn't find it, but it's 50dB & seems to work well out to 2GHz
<bvernoux>
and
<bvernoux>
JFW Industries Rotary Step Attenuator 0-70db 2.4GHz 50DR-068 SMA
<bvernoux>
this one is not standard
<bvernoux>
miek, yes I expect they have a good margin it is why I will check that with my VNA up to 6GHz ;)
<azonenberg>
so you can see the bottom side of the sma doesnt have enough clearance
<azonenberg>
but otherwise they fit great
<bvernoux>
ha nice
<bvernoux>
the case is 3D Printed ?
<bvernoux>
it look like quite "dense" and we do not really see layers like on a 3D Printed part
<Degi>
I think its sintered powder
<bvernoux>
ha yes we see a texture like powder
<bvernoux>
but very clean
<bvernoux>
also the PCB is RO4350B ?
<azonenberg>
the case is MJF which is a selective sintering process similar to SLS
<azonenberg>
but instead of rastering a laser beam, it inkjet deposits binders then uses a heat lamp to do the actual fusing
<azonenberg>
i've had better results with this than SLS
<azonenberg>
for parts that have high aspect ratio holes inside them
<azonenberg>
the actual PCB is RO4350B, these are nonfunctional dummies made from FR4
<azonenberg>
just for fit testing the enclosure
<bvernoux>
ok
<azonenberg>
but yes, FDM is not a process i am a fan of
<bvernoux>
I do not really see what is wrong in fact
<azonenberg>
the problem is that the SMA is supposed to be recessed into the shell so only the threaded part sticks out
<azonenberg>
the bottom side "legs" collide with the enclosure
<bvernoux>
ha ok
<azonenberg>
there's an opening but it's too small
<bvernoux>
are you sure it is not an issue with tolerance with the process ?
<bvernoux>
part is a bit smaller than what is expected on STL file ?
<azonenberg>
no i went back and checked
<azonenberg>
the cad model is wrong
<azonenberg>
the clearance is like half what it should be
<bvernoux>
ha ok
<bvernoux>
and how it is in hand ?
<bvernoux>
and also in the handle
<bvernoux>
I need to buy some probe handle I do not have any ...
<azonenberg>
it feels ok although i havent glued the pcb in, it fits a bit loosely on one side
<azonenberg>
so it can fall out if not glued
<azonenberg>
so i havent poked too much with it yet
<azonenberg>
it fits the bipod nicely although i only have one of them
<azonenberg>
i guess you missed when the enclosure prototypes came in
<azonenberg>
i actually made three of them
<azonenberg>
this one, one out of acrylic photopolymer that was much more expensive and also i didn't like the texture of as much, and one out of this material with thinner walls that felt too flimsy
<azonenberg>
i like the slightly rough texture, it sits better in the hand
<bvernoux>
yes rough texture is nice for hand I think too
<bvernoux>
the PMK 893-250-00T seems expensive
<bvernoux>
do you have found it at good price ?
<bvernoux>
I expect something like 10USD max for that as it is just a plastic stuff not movable ...
<azonenberg>
It's steel filled plastic and i was quoted $48 each for the positioners
<azonenberg>
Which is why i only included them on the pro edition probe. I can't afford them otherwise
<azonenberg>
I do plan to make my own 3d printed version cheaper at some point, but not for the KS
<azonenberg>
the pico branded version is much cheaper but i havent found a tetris sized one cheap
<azonenberg>
i also want to figure out something else, like a helping hand style ball joint, for more flexibility
<azonenberg>
the bipods are cheap but a bit limiting
<azonenberg>
incidentally lecroy sells the exact same bipod for $101 lol
<azonenberg>
at digikey. Tequipment has it for "only" $89.12
<azonenberg>
batronix.com looks to sell the PMK version for 27.50 EUR
<azonenberg>
plus VAT and shipping, but given the 250 USD shipping from PMK that might still be a better option
<bvernoux>
yes interesting
<bvernoux>
the Pico seems to be the best for price
<bvernoux>
but they are not compatible with your probe ?
<azonenberg>
the pico branded positioner you probably see for 25 USD or so is too tight
<azonenberg>
the tetris version is about 1mm wider opening
<bvernoux>
ha ok
<azonenberg>
it fit my older version probes
<azonenberg>
but not this one
<bvernoux>
The idea for Probe Positionner is it work for different stuff
<bvernoux>
like standard Oscilloscope Probe and your probe
<bvernoux>
it seems it is the case for PMK
<bvernoux>
Batronix sell it as 2-FUSS-TETRIS
<bvernoux>
for 27Euros
<bvernoux>
27.5Euros it is the one you say
<azonenberg>
yes
<azonenberg>
if it's specifically labeled as for the tetris series, it's the right one
<bvernoux>
it is crazy to see all those different price for the same things ;)
<azonenberg>
ok i just ordered a couple of the latest version of the shell
<azonenberg>
with the new cutout for the sma
<bvernoux>
ha great
<bvernoux>
what is the delay to receive it ?
<bvernoux>
as a nice update on the KS will be welcome ;)
<_whitenotifier-c>
[starshipraider] azonenberg pushed 1 commit to master [+0/-0/±1] https://git.io/Jf8SZ
<_whitenotifier-c>
[starshipraider] azonenberg c2dc538 - Updated probe shell design with notch for SMA connector bottom side pins
<azonenberg>
i should have it by the 27th or so. but i will probably do an update in a day or two with where i am now
<azonenberg>
showing the dummies
<bvernoux>
ha niec
<bvernoux>
nice
<bvernoux>
about 75% reached
<azonenberg>
Yeah
<bvernoux>
missing something like 1KUSD
<bvernoux>
in 17days it is clearly very good
<bvernoux>
with 17 days to go
<azonenberg>
yeah somebody bought the pro version today which helped bring it up a ton
<bvernoux>
yes it is clearly a very interesting package for the pro
<bvernoux>
for 550USD
<bvernoux>
I think COVID-19 does not help ...
<azonenberg>
Yeah agreed
<bvernoux>
I will have speak about that at my work but so far all is done in remote(@home)
<bvernoux>
It is very interesting to probe RF on circuit
<azonenberg>
I'm planning on doing at least one more spin after the KS to fine tune even more, but i need to characterize this rev and probably upgrade sonnet first
<azonenberg>
i think i can push this design a lot further than i'm running it now
<bvernoux>
I think your shall speak about that as there is no keyword for RF PCB probing...
<bvernoux>
For example to caracterize 2.4GHz Wifi/Bluetooth Antenna with a Spectrum Analyzer
<bvernoux>
as the probe go to 2GHz it is clearly usable for 2.4GHz for WiFi...
<bvernoux>
it is just an example
<bvernoux>
as it is often not convenient at all to check such small Antenna with UFL connector ...
<bvernoux>
It will be interesting to see the effect of probe on different Antenna too
<azonenberg>
this probe actually drops off a fair bit right past 2 GHz
<azonenberg>
i want to figure out why
<bvernoux>
in worst case it can be great to tune other stuff <2GHz ;)
<bvernoux>
2.4GHz will be nice anyway ;)
<bvernoux>
(I always want more ;)
<azonenberg>
lol who doesnt want more bandwidth
<Degi>
Just cut a SMA cable and remove some shielding and use that as a probe lol
<Degi>
50 ohm loading
* monochroma
has some plans for 60GHz testing in the near-ish future
<bvernoux>
monochroma, what are the plans for cheap 60GHz testing ;) ?
<monochroma>
less direct probing, but qualcomm makes some front end ICs for 60GHz WiFi applications
<bvernoux>
A fun experimentation to detect electronic devices ;)
<bvernoux>
Especially those who think COVID-19 comes from 5G tests ...
<bvernoux>
I love the reduction rate in % and not in dB ;)
<azonenberg>
wait what
<bvernoux>
they have forgotten to shield the bottom ;)
<azonenberg>
"avigation"
<miek>
that's because the covid towers are mounted high up
<bvernoux>
haha
<azonenberg>
"ecology: normal"
<azonenberg>
not sure what that is supposed to mean
<bvernoux>
haha
<bvernoux>
it means it is not too toxic ;)
<bvernoux>
it is not a good farady cage anyway
<azonenberg>
they have a plausible looking shielding plot at the bottom. this might actually be a reasonable price per square meter of rf shielding material if you wanted to build your own small faraday cage?
<bvernoux>
faraday
<azonenberg>
i havent looked into what it costs in bulk
<bvernoux>
but they missed the bottom ;)
<azonenberg>
i dont mean using it as is
<bvernoux>
where RF enter
<azonenberg>
i mean cutting it up and using it as a raw material
<monochroma>
knockoff faraday cage: Friday cage
<bvernoux>
ha yes maybe if it is not fake ;)
<azonenberg>
lol well thats a whole other situation
<bvernoux>
they say it is silver fiber anyway ;)
<bvernoux>
100%
<azonenberg>
i read it as silver plated cloth
<azonenberg>
but no idea how thick the silver is
<bvernoux>
yes so not 100% silver ;)
<bvernoux>
the graphic with attenuation
<bvernoux>
is not clear
<bvernoux>
does it means it attenuate -50dB ;)
<bvernoux>
not really
<bvernoux>
as we do not see with and without ;)
<miek>
i was thinking it could be neat for a collapsible dish antenna
<bvernoux>
I think the quality is not that good ;)
<bvernoux>
it seems they sell that against mosquito ;)