azonenberg changed the topic of #scopehal to: libscopehal, libscopeprotocols, and glscopeclient development and testing | https://github.com/azonenberg/scopehal-apps, https://github.com/azonenberg/scopehal, https://github.com/azonenberg/scopehal-docs | Logs: https://freenode.irclog.whitequark.org/scopehal
_whitelogger has joined #scopehal
juli965 has quit [Quit: Nettalk6 - www.ntalk.de]
<_whitenotifier-f> [scopehal] azonenberg pushed 1 commit to master [+0/-0/±4] https://git.io/JU9Pa
<_whitenotifier-f> [scopehal] azonenberg 1478a58 - Fixed namespace collision between PacketDecoder and Filter colors
<_whitenotifier-f> [scopehal-apps] tarunik commented on issue #180: Look into static analysis options - https://git.io/JU9PD
<_whitenotifier-f> [scopehal-apps] azonenberg commented on issue #180: Look into static analysis options - https://git.io/JU9Xs
<_whitenotifier-f> [scopehal-apps] azonenberg edited a comment on issue #180: Look into static analysis options - https://git.io/JU9Xs
Degi has quit [Ping timeout: 240 seconds]
<_whitenotifier-f> [scopehal-docs] azonenberg opened issue #19: Build doesn't depend on images - https://git.io/JU9XH
Degi has joined #scopehal
<_whitenotifier-f> [scopehal-docs] azonenberg pushed 1 commit to master [+4/-0/±10] https://git.io/JU91S
<_whitenotifier-f> [scopehal-docs] azonenberg 708a743 - Lots of updating of documentation, added graphics for new toolbar buttons
<_whitenotifier-f> [scopehal-apps] tarunik commented on issue #180: Look into static analysis options - https://git.io/JU91F
<_whitenotifier-f> [scopehal-apps] azonenberg commented on issue #180: Look into static analysis options - https://git.io/JU9MZ
electronic_eel has quit [Ping timeout: 258 seconds]
electronic_eel has joined #scopehal
<_whitenotifier-f> [scopehal-docs] azonenberg pushed 1 commit to master [+2/-0/±4] https://git.io/JU9yC
<_whitenotifier-f> [scopehal-docs] azonenberg 8af472d - Added trigger section to manual
<_whitenotifier-f> [scopehal-docs] azonenberg pushed 1 commit to master [+0/-0/±1] https://git.io/JU95Z
<_whitenotifier-f> [scopehal-docs] azonenberg 399b42e - Finished initial trigger documentation
azonenberg_work has quit [Ping timeout: 240 seconds]
azonenberg_work has joined #scopehal
Nero_ has joined #scopehal
Nero_ is now known as NeroTHz
juli965 has joined #scopehal
bvernoux has joined #scopehal
<_whitenotifier-f> [starshipraider] azonenberg pushed 1 commit to master [+0/-0/±2] https://git.io/JUHo0
<_whitenotifier-f> [starshipraider] azonenberg 3f5feeb - Updated MEAD enclosure for v0.5
<azonenberg> NeroTHz: ping
<NeroTHz> pong
<azonenberg> NeroTHz: did you see the data from the v1.2 probe?
<azonenberg> Curious to hear your thoughts
<miek> i just got one of those midi controllers that noopwafel mentioned a few months back (the behringer x-touch mini). it seems very nice so far, and control definitely works in both directions so you can sync up the state with a pc app
<NeroTHz> I did not I think, azonenberg
<azonenberg> NeroTHz: ok so to start, i set up one of my attenuator test boards as a test fixture with better matching than the old one
<azonenberg> https://www.antikernel.net/temp/921-across-open.jpg this is the pico probe
<azonenberg> and this is my probe
<azonenberg> slightly different grounding due to the different tip geometry but that's as reproducible as i could get. For this particular test, the SMA had a 50 ohm termination across it and I was measuring S11
<azonenberg> i also measured S21 across the 50 ohm termination at the other end of the fixture
<azonenberg> Does this look like a reasonable fixture?
<azonenberg> nice clean matched GCPW all the way up to the end, then you hit the matched 0402 pad (which while not perfect, shouldn't significantly impair my results)
<NeroTHz> I geuss it should be okay
<NeroTHz> hard to tell really, I don´t have all that much experience down at those frequencies
<azonenberg> yeah i know you don't do DC much :p
<azonenberg> So here's the S21 curves. Pink is the v1.2 prototype, red is my best simulation of it, blue is the proposed v1.3 design. Cyan is the Pico 921 probe that is my main competition now
<azonenberg> there's a huge dip on the pico probe at 3.25 GHz that doesn't make much sense to me but it's quite good until then
<azonenberg> My probe now has -3 dB bandwidth of about 5.5 GHz and the sim suggests i should be able to flatten the lower half of that a bit with my new v1.3 design
<NeroTHz> that is already very good
<NeroTHz> quite impressive really
<azonenberg> yeah for a passive probe
<azonenberg> What i'm less thrilled about is S11
<azonenberg> Same colors for each trace. you can see my probe pulls the DUT signal down -6 dB at only 2.5 GHz
<azonenberg> while the pico probe doesn't hit that until about 5.1 GHz
<azonenberg> the red trace is my best fit model to the tip, it's 3.21 mm of 240 ohm line with a 1.0 dielectric constant (air) for the exposed portion of the tip, then 5.42 mm of 200 ohm line with a 3.5 dielectric constant (slightly below FR408)
<azonenberg> Those are actual physical dimensions for the length of the probe body and tip, guesstimated dielectric constants, then impedances hand tweaked to fit the measured data as well as i could
<azonenberg> there's some loss somewhere, perhaps skin effect plating on the tip or somethign, causing the actual probe's tip resonance to have lower Q than the simulation but the dip is the right size and center freq
<azonenberg> meanwhile the Pico probe's resonance is somewhere past 6 GHz but has much higher Q. It's off scale on this graph but when it hits the 6 GHz limit of my VNA it's at like -25 dB S11 and still dropping
<azonenberg> I did a control measurement on the fixture with no probe and there was only like -1 dB of S11 at 6 GHz from conductor losses etc on the fixture, so at this level of analysis we can neglect fixture losses and treat that open as a ~perfect broadband reflector
<NeroTHz> I wouldn´t worry too much about S11
<azonenberg> why? don't we want the tip to be a perfect broadband reflector (zero loading) in the ideal case?
<azonenberg> and since we have to pull some signal off, we should have a flat S11 slightly below zero?
<azonenberg> (probing across an open not a 50 ohm line)
<NeroTHz> wait you mean S11 seen from an ´open´ you are probing?
<azonenberg> Yes
<azonenberg> That's S11 across an open
<azonenberg> I'm fairly sure this is a null due to the tip resonating and acting like a stub
<azonenberg> causing standing waves between DUT and the resistors
<azonenberg> It's a short length of high impedance line
<azonenberg> I'll be doing separate S11 and S21 measurements across the new load fixture in a bit. but these numbers don't look good
<azonenberg> Anyway, the blue trace is what i simulated from https://www.antikernel.net/temp/new-tip3.png which switches to a new, slightly shorter tip socket
<azonenberg> and removes a little more extra metal from the tip
<NeroTHz> ah okay was confused. It could indeed be the tip resonating. I think the wide spacing of the pico might help them there - you have huge ´loop inductance´ of your transmission line there which makes it look better? it´s hard to say really
<azonenberg> This should lower input C even further, and also shift the resonance up
<azonenberg> And yes actually , i was thinking the same thing. I tried using one of the longer ground accessories in the other ground socket and did indeed get much nicer looking S11 plots but S21 was degraded too
<azonenberg> At lower data rates, though, this is already looking very nice https://www.antikernel.net/temp/probe-eyes.png
<azonenberg> this is a 1.25 Gbps, 150 mV swing on each leg differential pair
<azonenberg> with my probe on RXP
<azonenberg> Top eye is the DUT signal with probe loading, bottom signal is seen through the probe
<azonenberg> after de-embedding for measured probe+cable s-parameters
<azonenberg> you can see some eye closure on the bottom signal due to noise in the scope frontend, which isn't surprising - the signal off the probe is 15 mV swing
<NeroTHz> yeah that is a small signal indeed
<NeroTHz> looks good
<azonenberg> this is a pretty weak signal compared to stuff like DDR3 or a normal SERDES link (I started out with 6 dB of loss from the actual source)
<azonenberg> this was probing my SFP+ test fixture which has a 6 dB resistive splitter in each leg between the optic and the test port, with the other leg looping back to the other optic so you can do mid-span probing
<azonenberg> I stuck my probe between the RXP GCPW and ground
<azonenberg> while simultaneously using the SMA on that same line to look at the signal before and after probe loading
<azonenberg> I think this shows that at least out to 1.25 Gbps the impact of the current probe design on a real DUT is fairly minimal
<azonenberg> And i fully expect if i was probing something more than 150 mV swing the eye would look a lot nicer
<azonenberg> I also did some time domain loading measurements of the same 8b10b test signal
<azonenberg> https://www.antikernel.net/temp/LeCroy--00050.png this is the stock 500 MHz passive probe the scope comes with
<azonenberg> Light pink/yellow are memory traces of the signal with no probe loading, pink is signal with probe loading, green is through the probe
<NeroTHz> damn
<azonenberg> https://www.antikernel.net/temp/LeCroy--00051.png this is a Pico TA061 1.5 GHz transmission line probe
<azonenberg> Which i considered my main competition originally. I've now greatly surpassed its performance by every metric i can measure
<azonenberg> they have huge peaking on rising edges, about +3 dB of gain relative to nominal at ~800 MHz, then it falls off like crazy after about 1.2 GHz
<azonenberg> And the loading is quite high for a transmission line probe
<azonenberg> https://www.antikernel.net/temp/LeCroy--00053.png This is a LeCroy ZS1500 1.5 GHz active voltage probe
<azonenberg> https://www.antikernel.net/temp/LeCroy--00054.png this is the Pico 921 probe, which has the lowest loading of everything I've tried on this signal. Pretty decent looking signal
<azonenberg> interestingly when i turn the scope gain up any more than this the frontend bandwidth on the scope falls apart. https://www.antikernel.net/temp/LeCroy--00055.png this is one tick more V/div
<azonenberg> This is 50 mV/div on 20x probe which is 2.5 mV/div actual voltage at the scope, so i guess i'm just maxing out the GBP of some frontend amplifier?
<azonenberg> at 100 mV/div (5 mV/div actual) it looked great
<azonenberg> anyway this is only a problem if I am probing ultra low level signals with a high attenuation probe so i'm not too worried
<azonenberg> The realistic use case for this is probing 1000base-X and DDR3 signals and those are a lot higher swing
<azonenberg> Moving on, https://www.antikernel.net/temp/LeCroy--00056.png is my v1.0 probe (kickstarter rev
<azonenberg> https://www.antikernel.net/temp/LeCroy--00058.png this is the current v1.2 prototype using the tip mounted ground
<azonenberg> https://www.antikernel.net/temp/LeCroy--00059.png and this is the same probe using the longer "leaf" ground instead
<NeroTHz> last one 404´s
<azonenberg> reuploaded
<NeroTHz> huh interesting that with the long leaf ground it almost looks like a slew-rate limited signal
<NeroTHz> with the trapezoidal edges
juli965 has quit [Quit: Nettalk6 - www.ntalk.de]
<azonenberg> NeroTHz: it's probably the additional L of the long ground
<azonenberg> also i just got off the phone with an apps engineer at lecroy
<azonenberg> The calibration report for this scope has bandwidth measured at 1000, 100, 10 mV/div
<azonenberg> and the datasheet guarantees it will meet the advertised bandwidth down to 5 mV/div
<azonenberg> When you go below there bandwidth is unspecified. So basically i am indeed hitting frontend bandwidth limits at 2.5 mV/div
<azonenberg> Good to know
<azonenberg> Also just ordered the v1.3 probe board
<bvernoux> do you have picture of the v1.3 ?
<bvernoux> 3D render ...
<bvernoux> ha yes very compact and small
<bvernoux> I hope it will be the final version which will be nice up to 5GHz ;)
<azonenberg> I hope so, lol
<bvernoux> finally there is only 3 resistor now ?
<bvernoux> you can checked that with SOnnet simulation ?
<azonenberg> Yeah. 3 resistors 200-200-50
<bvernoux> so it is cheaper ;)
<azonenberg> My modeling right now is a combination of vendor s-parameters, sonnet field solver geometry, and simplified ideal netlist simulation
<azonenberg> If i wanted absolute max fidelity, modelithics has full sonnet *geometry* models of these resistors
<azonenberg> including the terminals, resistive element, etc
<azonenberg> to model things like capacitive coupling between the resistive element and the surrounding pcb geometry
<bvernoux> what is hard is to have a very flat response ;)
<azonenberg> But i doubt those models are cheap :P
<bvernoux> as in RF usually we want to filter a band with a passband ...
<azonenberg> and they probably need sonnet gold or pro to work given the level of detail in them
<azonenberg> Yeah
<azonenberg> broadband response is hard compared to being flat within one passband
<bvernoux> yes
<bvernoux> especially it is passive ;)
<bvernoux> you could be flat with some filter but at the end the signal will be very attenuated ;)
<azonenberg> Yeah
<azonenberg> Also, the other reason is that at this point i dont think the resistor model is my limiting factor anyway
<bvernoux> It is why I like this idea to do a nice passive probe
<azonenberg> i think the limit to simulation fidelity is now my tip model
<bvernoux> Yes I suspect also the resistor model is validated and the last things is the tip ;)
<bvernoux> or maybe optimize a bit the connector to PCB transition but IIRC you already have done that
<azonenberg> i originally optimized it for 50 ohm impedance
<azonenberg> this turned out to be the wrong choice
<azonenberg> higher is actually better so i pulled all the ground away from it
<bvernoux> yes here it shall be optimized for 500Ohms ;)
<azonenberg> yeah except it will be very hard to get the impedance that high
<bvernoux> 450Ohms in fact ;)
<azonenberg> it looks like its something around 250 right now
<bvernoux> yes it is why it is a very interesting challenge
<bvernoux> I have not found any paper about designing a good passive probe
<bvernoux> especially to exceed 2GHz BW and even more here the objective is 5GHz or more ;)
<bvernoux> the more the better ;)
<bvernoux> 5GHz is a good number to use the probe on USB 3.x line ;)
<bvernoux> anyway for USB 3.x probing I think the must is a diff probe
<bvernoux> it is an other challenge doing a very nice active diff probe up to 10GHz ;)
<azonenberg> Yes i want to build an active diff probe too
<azonenberg> we'll start low and then i'll push it as far as i can
<azonenberg> I want to do a solder in passive probe too
<azonenberg> i think i can remove the tip resonance and push that to 5+ GHz easily
<bvernoux> I'm waiting Qucs-RFlayout new version to test that ;)
<bvernoux> with 3D solver ;)
<bvernoux> what do you mean by solder in passive probe ?
<bvernoux> azonenberg, Do you have checked this project https://github.com/jankae/VNA2 ?
<bvernoux> it is clearly the best open source VNA which shall be very soon available
<bvernoux> I'm pretty sure we can improve lot of things without breaking too much the actual design
<bvernoux> as this version is intended to be compact at working from 1MHz to 6GHz and which is quite small about 10cm x 10cm
<azonenberg> bvernoux: the basic idea is a flex pcb with a long "tail" that solders into the DUT, has the same 3 resistors on it but a shorter stub between them and the DUT
<bvernoux> very far from my HP VNA but the performance so far are still far from my HP VNA which reach more than 100dB Dynamic Range from 30Khz to 6GHz
<azonenberg> then a sma connector at the other side
<bvernoux> ha yes nice idea
<azonenberg> i plan to make a differential version of the same thing suitable for using with an active diff probe too
<bvernoux> yes it is clearly something lot of guys want for a good price
<bvernoux> as they are crazy expensive ...
<azonenberg> the solder in version i think will easily hit 6+ GHz bandwidth
<azonenberg> and cost only a couple of dollars
<miek> i think it's worth looking closely at the swappable heads for the top end active probes, as far as i can tell they're a resistive probe on their own
<azonenberg> maybe $20ish
<azonenberg> miek: yes i think they are. But they're $$$$
<bvernoux> azonenberg, yes the solder in version is clearly a great things
<azonenberg> If i pick up a lecroy active diff probe i am seriously considering trying to design my own tips for it
<bvernoux> It is just amazing as no anyone have done cheap probe for scope which exceed GHz BW
<miek> there's a lot of material out there without buying one. keysight even gives out spice models for theirs
<bvernoux> It will be a game changer for the industry to provide that for /100 price ;)
<bvernoux> Lecroy, R&S, Tek and other will cry ;)
<azonenberg> miek: oh interesting. anyway, one thing at a time
<azonenberg> the handheld browser probe is approaching limits of how far i can push it for the moment
<azonenberg> the next step will be the solder in single ended probe tip
<azonenberg> Then an active diff tip
<azonenberg> active diff amplifier + passive diff tip*
<bvernoux> yes a pluggable diff tip
<bvernoux> like Tek ;)
<bvernoux> I suspect others do similar things with flexpcb for the tip which contains 2 resistors ...
<azonenberg> It's more than two
<miek> afaik keysight's are all two per channel, and an extra axial one as close as possible to the probe point if you're soldering in long leads
<bvernoux> miek, do you have such probe ?
<bvernoux> I have never see any of those ultra expensive diff probe for >5GHz ;)
<miek> nope, not yet :p
<bvernoux> I have already seen some GHz LA probes for DDR3/4 ... ;)
<bvernoux> but it is not the same stuff
<azonenberg> the lecroy solderin diff tip for their 4-6 GHz probes has two SMT resistors plus two axial lead ones
<azonenberg> and i think i saw one, cant remember if keysight or lecroy, that had a differential resistor too between the two legs
<azonenberg> also interesting to note that this is a flex-rigid where the tip is actually a rigid board
<azonenberg> whereas in my design the tip is flex all the way to the end and uses castellations to mount to the DUT
<miek> i think i saw that diff resistor on the lecroy one you linked the other day
<bvernoux> ha nice
<bvernoux> it look like they have embedded a filter ;)
<bvernoux> on PCB
<bvernoux> which look like a LED ;)
<azonenberg> interesting they have shield between the two diff legs
<azonenberg> so they're not tightly coupled
<azonenberg> it's two independent GCPWs basically
<bvernoux> it seems to be an interdigital filter ;)
<bvernoux> miek, do you have high resolution of that picture ?
<bvernoux> I love those filters
<bvernoux> It is why I would like a CNC to do such things ;)
<bvernoux> some nice cavitity filters
<bvernoux> cavity
NeroTHz has quit [Read error: Connection reset by peer]
<bvernoux> the performance is amazing with cavity filters
<miek> the white parts are supposed to be 25kohm resistors, but i don't know what other wizardry is in there :p
<miek> hard to find really good zoomed in shots, there's another angle in here though: https://www.keysight.com/us/en/assets/3119-1107/data-sheets/MX0023A-InfiniiMax-25-GHz-RC-Probe.pdf
<bvernoux> miek, if you zoom on white part you see it is like a stripline interdigital filter ;)
<bvernoux> woo 25GHZ BW ;)
<bvernoux> totally crazy ;)
<bvernoux> I confirm it is an interdigital filter
<bvernoux> with a good zoom we see it
<bvernoux> but even smaller ;)
<bvernoux> it is ultra wideband
<azonenberg> sooo i think tonight i'm going to start design on the solder in probe tip
<azonenberg> Thinking about how i want to actually design it
<azonenberg> in particular, my previous solder in tips used castellations and had contacts on both top and bottom side
<azonenberg> i'm not sure if that makes the most sense?
<azonenberg> because castellations have a minimum size due to the drill
<azonenberg> i wonder if for smaller stuff it might make more sense to have a top-side-only contact that you then solder with ultra-thin copper wire to your test point
<azonenberg> so basically you'd kapton tape the probe very close to the DUT then wire it in place
<bvernoux> haha I'm reading spec of Keysight 110GHz scope
<bvernoux> Every UXR-Series includes female-to-female connector savers to help protect the unit’s primary bulkhead connector from damage.
<bvernoux> I imagine you kill one of this connector it cost the price of a luxurious car to replace it ;)
<azonenberg> Lol
<bvernoux> they are using Indium Phosphide ;)
<azonenberg> yes i would imagine
<azonenberg> do you know how much a 1mm connector costs just by itself?
<azonenberg> The only one in stock at digikey is $805
<azonenberg> just for the connector, not counting the cost of replacing it and recalibrating everything
<bvernoux> haha
<bvernoux> yes crazy
<bvernoux> on my side the limit is 2.92mm connectors ;)
<miek> i'll be staying down at "DC" for a while yet :)
<miek> re: probe wiring, that sounds good to me
<bvernoux> what will be fun in few years is ADAS stuff ;)
<bvernoux> the next gen
<azonenberg> Obviously the closer you can get it to the DUT, the better the response will be
<azonenberg> but i think we can basically eliminate the tip resonance - or shift it well past 10 GHz to the point that it won't matter
<azonenberg> then use the same, tested attenuator design
<bvernoux> yes probe outside 5GHz are interesting not for scope but for SA ;)
<azonenberg> well i eventually want to be able to do in system probing of 10GbE
<bvernoux> as anyway it is ultra expensive to have scope > 5GHz BW ;)
<azonenberg> For testing i'll make the board on OSHPark which will be fun because they do not have stiffeners
<azonenberg> So getting an edge launch SMA onto flex will be interesting
<bvernoux> they can sponsor you for that
<azonenberg> what i might do is just cut a random scrap of FR4 and shove it under the connector
<bvernoux> to do free PCB
<bvernoux> they are very nice for open source RF stuff
<bvernoux> they have sponsored my latest TRL board
<bvernoux> else it was costing >100USD in PCB
<bvernoux> for 3 PCB so it start to be expensive for tests especially when you know you will need to do 3 or 4 revisions (or even more if you want to push it to the edge and add other stuff)
<bvernoux> Trump have the COVID ;)
<bvernoux> what a crazy guy it is first time I see a present so carzy
<bvernoux> crazy
<bvernoux> president
<bvernoux> not present ;)
bvernoux1 has joined #scopehal
bvernoux has quit [Ping timeout: 240 seconds]
electronic_eel has quit [Ping timeout: 246 seconds]
electronic_eel has joined #scopehal
<azonenberg> bvernoux1, miek, lain, monochroma: https://www.antikernel.net/temp/flexprobe-01.png
<azonenberg> proposed probe design so far
<azonenberg> about 150mm / 6 inches long
<azonenberg> initial prototype is bare GCPW, but i like the way lecroy "sleeved" theirs so i might try and figure out a way to put some kind of protective shell over it
<azonenberg> actually thinking about the impedances more, it might make sense to do non-grounded CPW... hmm