azonenberg changed the topic of #scopehal to: libscopehal, libscopeprotocols, and glscopeclient development and testing | https://github.com/azonenberg/scopehal-apps | Logs: https://freenode.irclog.whitequark.org/scopehal
Degi_ has joined #scopehal
Degi has quit [Ping timeout: 245 seconds]
Degi_ is now known as Degi
Bird|ub3rghosted has joined #scopehal
Error_405 has joined #scopehal
Bird|ub3rghosted is now known as Bird|otherbox
balrog_ has joined #scopehal
noopwafe1 has joined #scopehal
deltab_ has joined #scopehal
monochro1a has joined #scopehal
vup2 has joined #scopehal
electronic_eel_ has joined #scopehal
Ekho has quit [Disconnected by services]
electronic_eel has quit [*.net *.split]
Error_404 has quit [*.net *.split]
Bird|ghosted has quit [*.net *.split]
balrog has quit [*.net *.split]
monochroma has quit [*.net *.split]
noopwafel has quit [*.net *.split]
vup has quit [*.net *.split]
deltab has quit [*.net *.split]
balrog_ is now known as balrog
monochro1a is now known as monochroma
Ekho has joined #scopehal
noopwafe1 is now known as noopwafel
<azonenberg> S11 and Zin measurements from the first four AKL-PT2 units
<azonenberg> A68-001 is quite the outlier, going to remeasure it and see if it was a bad soldering job or maybe if something is wrong with it somehow
<azonenberg> I'm switching to a different serial number scheme to fit the very tiny space available on the underside of the FR4 stiffener (the only part of the probe with enough space for a S/N)
<azonenberg> A = 1 year since 2020, 68 = day of year, then 3-digit serial within that day
<noopwafel> that is a pretty fun outlier
<azonenberg> This is measured across an open circuit
<azonenberg> what's interesting is the frequency of that null, and the fact that all probes have a null there
<azonenberg> but A68-001 is much worse
<azonenberg> that frequency lines up nicely with a reflection off the SMA
<azonenberg> So I think it's a mismatch there
<azonenberg> I did use a lower grade terminator for measuring 001 then switched to a nice one. I dont think I'd get 0.8 dB more reflection through two resistors from a cheap terminator but i'll definitely remeasure
<azonenberg> I'll also get a lot more data once my next batch of 25 SMAs arrives later in the week
<azonenberg> when i have 20+ probes worth of measurements I can draw much more conclusions about yield, consistency, etc
<noopwafel> I look forward to more stories, if nothing else this is all v.educational :)
deltab_ is now known as deltab
<azonenberg> also i just compared all of the PT2's to the pico 921
<azonenberg> even A68-001 has lower loading than it
<azonenberg> Which makes sense because i don't have the stub from the tip needle
<miek> azonenberg: could i get a copy of the s1ps for those? i want to have a play around with doing transforms on the data
<miek> thanks
vup2 is now known as vup
<azonenberg> https://www.antikernel.net/temp/pt2-s21-2.png and this time A68-004 is the outlier (on S21), dipping way low starting around 3.5 GHz
<azonenberg> sorry I meant A68-002
<azonenberg> I can't fully assemble the rest of the probes until I get the SMAs (current ETA is Friday) but I might do SMT-only assembly on a panel or two tonight or tomorrow
juli968 has joined #scopehal
bvernoux has joined #scopehal
<azonenberg> Preliminary AKL-PT2 characterization data
<azonenberg> Will be filled out more as I get more probes to test
<azonenberg> Leaving rise/fall time numbers blank for now until i get the new scope
<azonenberg> hoping it will come in time
asy__ has quit [Ping timeout: 272 seconds]
asy_ has joined #scopehal
<azonenberg> bvernoux, noopwafel: re-measured A68-001. Not sure why it was so bad the first time
<azonenberg> S11 is almost perfectly flat from DC to 6 GHz. This plot is 1 dB from top to bottom, 0.2 dB/div
<azonenberg> those "huge" ripples around 3 GHz are actually about half a dB p-p lol
<azonenberg> https://www.antikernel.net/temp/pt2-vs-921-zin.png here's the same plot with the Pico 921 thrown in for comparison's sake
<azonenberg> in green
<bvernoux> ha nice let's see the png
<bvernoux> some small ripples but very flat
<bvernoux> I can test it up to 15GHz if you want
<bvernoux> with my RF siggen + SA
<bvernoux> it is good also to check with real signals
<bvernoux> even if the must have is VNA
<bvernoux> ripples could be related to Wifi/BT at 2.4GHz
<bvernoux> but they seems very small like less than 0.5dB it is hard to see with the picture
<bvernoux> it is strange as the red ones have more ripple
<bvernoux> If i understand it correctly A68-001 to 004 are just 4 probes built with exactly same material to compare the stability on 4 build ?
<azonenberg> Correct
<azonenberg> This is S11 just to be clear
<bvernoux> yes S21 is the most interesting
<azonenberg> 002 is the outlier, not yet sure why. It fails the 6 GHz bandwidth requirement (dipping below -23.5 GHz from about 4.5 to 5 GHz then a few times after that)
<azonenberg> but 001, 003, 004 are all >6 GHz probes with fairly smooth rolloff
<bvernoux> it is very flat too
<bvernoux> the blue is a bit off compared to other it is strange but the scale is very small too
<azonenberg> I'm going to be testing a lot more units in a few days
<azonenberg> Friday I have 25 more SMAs coming
<azonenberg> I stopped at four this time because that's all i had connectors for
<azonenberg> but i have 241 more bare PCBs
<azonenberg> I'm planning to do SMT assembly of two panels today or tomorrow then on friday put the SMAs on them and measure
<azonenberg> that should give me enough units to open initial sales, as well as enough test data to write a datasheet with reasonable limits
<azonenberg> https://www.antikernel.net/temp/akl-pt2-manual.pdf is the draft of the datasheet so far, if you have any suggestions?
<azonenberg> it's modeled on the AKL-PT1 datasheet, section 7/8 are blank awaiting more test data
<azonenberg> 6 is missing some data
<miek> what is the test setup for these measurements?
<azonenberg> Descriptions of the test setup is one of the sections of the datasheet I haven't written yet. actually talking to an artist to go draw up some diagrams as they'll be more clear than photos
<azonenberg> the tl;dr is that the S11 curves are measured with the probe soldered across the end of a 6mm long open circuit line
<azonenberg> and the S21 curves are measured with the probe soldered across a 50Ω termination
<azonenberg> I will also have typical rising edge waveforms with cables de-embedded. The holdup there is that I need a scope with >6 GHz bandwidth to accurately measured that :p
<miek> and the cal ref. plane is at a connector on that open line?
<azonenberg> No
<azonenberg> well
<azonenberg> calibration plane is at connector but i did an extension to the end of that open
<azonenberg> so the impedance numbers are based on the shifted ref plane
<miek> aha ok
<azonenberg> For the S21 measurements since the fixture i use there is a bit longer, I made a sonnet model of the fixture
<azonenberg> tuned some numbers to match VNA measurements of a similar thru line
<azonenberg> and de-embedded that s2p from the measurement
<azonenberg> plus a fixed ref plane shift to remove the electrical length of the SMA
<azonenberg> So basically in both cases the measurement plane should ideally be at the tip of the probe
<miek> for the s11 measurements, is the probe still setup with some length of coax going back to port 2?
<azonenberg> No. There's a 50 ohm termination across the SMA and port 2 of the VNA is floating
<azonenberg> also
<azonenberg> old and new measurements of probe 001
<miek> ah ok. hmm. i'm trying to make sense of the small evenly spaced ripples
<miek> nothing in the test setup seems long enough to make ripples with that small of a frequency spacing
<azonenberg> Best guess is reflections between the resistors
<azonenberg> the 0402s are "touching" but there's actually ~0.6mm of electrical length between the nichrome elements
<miek> https://blog.teledynelecroy.com/2020/05/reading-s-parameters-ripples.html this & the rest of the series are really good posts btw
<azonenberg> Will have to read that
<miek> i've also been reading Joel Dunsmore's book again after he was on the amp hour, there's some really cool stuff in there about time domain gating & automatic fixture removal
<bvernoux> the link on lecroy article about ripple is interesting
<bvernoux> but here ripples are very localized between 2.4 to 2.6GHz
<bvernoux> on S11
<bvernoux> on S21 yes it like like what is described by lecroy article
<bvernoux> especially the ripples are very small
<bvernoux> about +/0.5dB
<miek> here's a little experiment with just gating the first 5ns on the probe s11: https://i.imgur.com/eC85MTt.png
<azonenberg> miek: is that the old or new 001?
<bvernoux> anyway for such probes which are unshield and quite long I doubt it is possible to do something better
<miek> that's the old one
<miek> (the one from a few hours ago)
<azonenberg> yeah
<bvernoux> i see in spec the PCB length is 150mm
<azonenberg> If I had a higher bandwidth vna or tdr
<azonenberg> you should see three distinct reflections in the time domain on s11
<azonenberg> one off each resistor
<azonenberg> And yeah 5ns of gating should be more than enough given that the probe is only 925ps long end to end
<bvernoux> I think about characteristics you could round the value to integer
<azonenberg> so 2ns gives time for a full end to end reflection
<bvernoux> like Typ -21.95 -> 22 ...
<bvernoux> -22
<azonenberg> bvernoux: these numbers are rough. in the final datasheet the typical number will be averaged from the first production batch
<azonenberg> min/max will be min/max of the production batch plus a bit of safety margin, after discarding any probes i declare to be out of spec
<bvernoux> yes it is hard to provide accurate number until you have done check on lot of probes and quote min/max
<azonenberg> That's how i did spec for the PT1
<azonenberg> Next weekend I will have 20+ probes worth of characterization data
<azonenberg> The numbers i have there are averaged from three probes :p
Stary is now known as stary-but-differ
stary-but-differ is now known as Stary
sam210723 has quit [Read error: Connection reset by peer]
sam210723 has joined #scopehal
bvernoux has quit [Quit: Leaving]
<azonenberg> and hmm, i never figured out wtf was going on with multiple clFFT filters running in parallel
<azonenberg> i think for now as a workaround i'm just going to add a hint to each class that uses clFFT
<azonenberg> and have the scheduler serialize them
<azonenberg> which will hopefully prevent whatever conflict i'm hitting
electronic_eel_ has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
electronic_eel has joined #scopehal
<Famine> azonenberg, started to simulate with non-ideal resistors and it is amazing how much gate ringing i get
juli968 has quit [Quit: Nettalk6 - www.ntalk.de]