<azonenberg>
So i have a couple of trivial measurements left to convert to the new filter model (over/undershoot and peak to peak)
<azonenberg>
then all of the metrics on eye patterns
<azonenberg>
the thing is, eye width/height etc don't make sense as vector measurements
<azonenberg>
So how would you suggest implementing that?
<azonenberg>
one idea: have some kind of flag for a filter that says "don't show the waveform, only enable stats"
<azonenberg>
And then have the output be a vector with a single element
<azonenberg>
but part of the point of unifying all of this stuff was to avoid too much special casing
<lain>
azonenberg: wait why can't you just do eye width/height as scalar measurements?
<azonenberg>
lain: so i could do them as statistics
<azonenberg>
the problem is that then i'd have to add more rows to the table
<azonenberg>
if you have a lot of eye parameters thats lots of wasted space in the table that isn't meaningful for other channels
<azonenberg>
the basic structure i have right now is channels in the columns and statistics in the rows
<azonenberg>
"eye bit rate" is a plausible stat, it just would mean lots of rows in the table
<azonenberg>
with lots of blank entries under other columns
<azonenberg>
but i guess if you have several eyes it might actually work out ok?
<azonenberg>
i guess then that means i need to implement support for adding stats to the table
<azonenberg>
because right now i hard code min/max/avg as the only entries and there's no way to turn rows on or off
<azonenberg>
long term plan is to have average the only one shown by default and others enabled as needed
<azonenberg>
lain: oh wait there's another problem with this idea
<lain>
ono
<azonenberg>
one of the points of unifying measurements and math etc was to take advantage of the existing infrastructure for configuring properties and inputs etc
<azonenberg>
stats were supposed to be simple reduction functions with no inputs
<azonenberg>
But then how do i handle "eye height" on say a PAM4 signal?
<lain>
is eye height on a PAM4 actually a vector?
<lain>
or do you take min()
<azonenberg>
that's actually what i was thinking
<azonenberg>
there's a few options
<azonenberg>
one is to have "eye height" a scalar measurement of a single eye
<azonenberg>
that takes a parameter specifying the voltage to measure at
<azonenberg>
The other is a vector with one entry per opening
<azonenberg>
if it's a vector it makes sense to use stats for it
<lain>
I guess if you allow parameterizing stats, then you have a UI problem of where to display those parameters
<azonenberg>
yes, and configure them etc
<azonenberg>
i really want to keep stats simple and have all of the config in the filter graph
<azonenberg>
Related: i don't currently have proper support for multilevel eyes anyway - this might be a good excuse to do that
<azonenberg>
for the moment i do clock recovery on MLT3 by picking either the upper or lower transition as a trigger point and having the pll free-run between other values
Degi has quit [Ping timeout: 272 seconds]
Degi has joined #scopehal
<azonenberg>
Hmm
<azonenberg>
lain: so i have another idea
<azonenberg>
i wanted to do bathtub curves of eye patterns
<azonenberg>
basically extract a 1D slice of an eye and plot it as frequency vs time within the UI
<azonenberg>
what if i have a single "width" measurement that can be applied to an arbitrary X or Y axis oriented slice across an eye pattern?
<azonenberg>
or wait no even that will still be a problem
<azonenberg>
it would work for width
<azonenberg>
but a Y axis bathtub plot across a MLT3 eye has two regions to measure width of
<_whitenotifier-9>
[scopehal] azonenberg pushed 1 commit to master [+0/-0/±1] https://git.io/Jflea
<_whitenotifier-9>
[scopehal] azonenberg 3785601 - LeCroyOscilloscope: fixed bug in IsChannelEnabled() causing digital channels to not hit the cache
<azonenberg>
huh so there's a weird performance bug (regression): testing with ethernet on channels 1-4 of my waverunner
<azonenberg>
if i evaluate ch1-ch2 it's fine
<azonenberg>
if i do ch3-ch4 it's fine
<azonenberg>
if i enable BOTH framerate slows to a crawl for reasons that aren't obvious
<azonenberg>
more interesting is that the refreshing of the filters happens super fast, just a few ms
<azonenberg>
so the slowdown is somewhere else
<azonenberg>
i really need a better integrated performance monitoring system to track where all of my time is going
<azonenberg>
innnteresting it seems like any time i have two decodes enabled it slows down horribly
<azonenberg>
so... with no decodes, RefreshAllDecoders) takes 17 us, C1-C2 takes ~20 ms
<azonenberg>
when i enable the cdr pll it's a little bit slower but not much
<azonenberg>
... but despite the refreshing happening super fast the GUI slows to a crawl
<_whitenotifier-9>
[scopehal] azonenberg pushed 1 commit to master [+0/-0/±1] https://git.io/JfleA
<_whitenotifier-9>
[scopehal] azonenberg 4afdc86 - LeCroyOscilloscope: don't lock m_mutex during IsChannelEnabled() until we're sure that we actually need to send a command to the hardware
<_whitenotifier-9>
[scopehal] azonenberg pushed 2 commits to master [+0/-0/±3] https://git.io/JflvB
<_whitenotifier-9>
[scopehal] azonenberg dbe600f - ClockRecoveryDebugDecoder: Y axis is properly displayed in ps
<azonenberg>
Off to bed, then bathtub plots hopefully tomorrow morning before work
<azonenberg>
so much for doing probe design and AFE testing today
futarisIRCcloud has quit [Quit: Connection closed for inactivity]
<_whitenotifier-9>
[scopehal-apps] azonenberg labeled issue #91: Protocol decoder dialog should not freeze entire UI when active - https://git.io/Jfl2h
<_whitenotifier-9>
[scopehal-apps] azonenberg opened issue #91: Protocol decoder dialog should not freeze entire UI when active - https://git.io/Jfl2h
<_whitenotifier-9>
[scopehal-apps] azonenberg labeled issue #91: Protocol decoder dialog should not freeze entire UI when active - https://git.io/Jfl2h
bvernoux has joined #scopehal
<bvernoux>
hello
<bvernoux>
haha some Ebay guys are crazy ;)
<bvernoux>
AGILENT HP DSO80204B 2GHz, 4 canaux, 40 GSA/S Digital Oscilloscope InfiniiVision => for 86USD ;)
<azonenberg>
what's wrong with it?
<bvernoux>
nothing it is fully working
<bvernoux>
anyway the bid end in 2 days
<bvernoux>
and the price will fly
<azonenberg>
oh
<bvernoux>
there also amazing Lecroy 6GHz BW 40GSPS ;)
<bvernoux>
the return ;)
<azonenberg>
i thought thatwas the buy now price
<azonenberg>
or it had actually sold for that much
<bvernoux>
tons of crazy price for different hw ;)
<azonenberg>
yeah that doesnt sound sketchy at all
<bvernoux>
ha yes
<bvernoux>
all those are fake too good to be true ;)
<bvernoux>
even at 6KUSD it was a nice price
<bvernoux>
for something which was > 50KUSD
<bvernoux>
ha it was for part not working ;)
<miek>
most saturdays this stuff goes up. they scrape a bunch of old listings for high value stuff, find an account with high feedback to take over (credential stuffing or something), and fill it with listings that say email us to get this absurd buy-it-now price
<Degi>
how do they take an account over?
<azonenberg>
the first red flag is an account that sells stamps suddenly switching to test equipment
<azonenberg>
:p
<Degi>
ah yes
<monochroma>
"julies-collectables"
<bvernoux>
yes the name is funny ;)
<bvernoux>
they are specialized to sell stamps and they have >50K USD Scope to sell for <300USD ;)
<bvernoux>
also the Flir T420 High performance Infrared Camera 320x240 Resolution, Wi-Fi, Bluetooth
<bvernoux>
for 102GBP ;)
<bvernoux>
such stuff cost >7KUSD
<monochroma>
to be fair, i have bought some odd server equipment from som collectables sellers... but i was not happy when they shipped a 2U server in nothing but USPS boxes taped around it (and of course it came extremely beaten up...)
<bvernoux>
ha bad
<bvernoux>
yes maybe some guys buy those stuff for nothing and as they do not know what is it they sell it at low price ...
<bvernoux>
but anyone not stupid do a check on Ebay or other to check price for similar item;)
<bvernoux>
what is overpriced is Spectrum Analyzers
<azonenberg>
miek: my other favorite ebay kinda-scam involves selling an item you dont actually have, then buying it and shipping it a significant time later
<bvernoux>
especially the E440x
<azonenberg>
but you ship a package with either nothing or a worthless item first, supply tracking
<azonenberg>
hoping the user will submit feedback etc before the package comes
<azonenberg>
then the package arrives and it's not what you ordered, it has the junk and a little note saying "here's a free gift for the inconvenience. the real item is coming later"
<azonenberg>
later meaning after the dispute resolution window has ended
<bvernoux>
monochroma, and the server was working or there was some missing stuff in it ?
<azonenberg>
i have some refactoring to do to make eye/bathtub handling share some common code regarding window resizing etc
<Degi>
neato
<azonenberg>
and i have to do vertical bathtubs too
<azonenberg>
and probably make bathtubs use a log scale for the y axis? that make more sense?
<Degi>
Is that like a hisstogram?
<azonenberg>
a bathtub is basically a 1D slice through an eye pattern
<Degi>
Maybe switchable between lin and log
<miek>
azonenberg: i almost got bit by something like that - bought something from china that was supposedly in the UK, i knew it wasn't so expected it to take a while to arrive. what i didn't realise was that the dispute window started at the expected delivery date D:
<azonenberg>
Yeah
<Degi>
oof
<azonenberg>
that's what they want
<azonenberg>
I filed a claim when the fake package showed up
<azonenberg>
got a refund
<azonenberg>
and was somewhat surprised when the actual item came a month later
<Degi>
lol
<Degi>
Is that 10G ethernet?
<azonenberg>
No that's 100M ethernet
<Degi>
Ah
<azonenberg>
i dont have a scope fast enough for 10G
<Degi>
Ah yes 10GS/s...
<azonenberg>
and while i could do 1Gbase-X, i dont have a probe setup on one of those handy
<_whitenotifier-9>
[scopehal] azonenberg pushed 1 commit to master [+2/-0/±7] https://git.io/Jflwn
<miek>
i wish ebay would filter this scam, it's the same template every time. it really messes with my test equipment searches :'(
<bvernoux>
yes huge scam
<sorear>
we call it "just-in-time ordering"
<Degi>
Kinda reminds me of oil traders with no handling capacity
<Degi>
Lol I just noticed that that eye diagram is 3 level...
<azonenberg>
Degi: yes thats why i'm using it as a test case
<bvernoux>
On my side I'm finishing my Amazon seller ;)
<azonenberg>
i dont have any PAM4 signal sources to test with but MLT3 will do fine for this
<bvernoux>
it is so complex just to sell few HW boards
<bvernoux>
btu advantage is to use Amazon to ship boards ...
<bvernoux>
as in France it is crazy complex & time consuming to ship things ...
<Degi>
Huh?
<Degi>
Hm around here you can just print a label, stick it onto a package and put it into a box somewhere, is it much different in france?
<bvernoux>
yes here for >40Euros ;)
<bvernoux>
shipping is more expensive that what is in the box a real joke
<bvernoux>
or you shall go to post office and pay >20euros for small packet to send in France or EU ...
<Degi>
Oh geez
<Degi>
Idk within germany I can send 2 kg for like 5? € with tracking
<bvernoux>
it is the price to send a letter here ;)
<Degi>
huh, its like 80 cents or something here...
<azonenberg>
meanwhile first class mail postage in the US is 55 cents for a letter
<bvernoux>
just to post a basic letter called green it is 0.97Euros here
bvernoux has quit [Quit: Leaving]
<_whitenotifier-9>
[scopehal-apps] azonenberg opened issue #92: Add support for "dense" waveform storage format (no explicit offset/len) to reduce size of analog captures - https://git.io/JflHo
<_whitenotifier-9>
[scopehal-apps] azonenberg labeled issue #92: Add support for "dense" waveform storage format (no explicit offset/len) to reduce size of analog captures - https://git.io/JflHo
<_whitenotifier-9>
[scopehal-apps] azonenberg edited issue #92: Add support for "dense" waveform storage format (no explicit offset/len) to reduce size of analog captures - https://git.io/JflHo
_whitenotifier-9 has quit [Ping timeout: 260 seconds]