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
maartenBE has quit [Ping timeout: 256 seconds]
maartenBE has joined #scopehal
<azonenberg> v1.3 probe prototype came in
<azonenberg> Waiting for it to cool off then i can test :D
<azonenberg> Holy moley
<azonenberg> i'm... very pleased with this rev
<azonenberg> i think i need to tone down the filter a little bit, it's too strong for the newly less-resonant tip socket
<azonenberg> but seriously, WOW
<azonenberg> lain: I can't even characterize the -3dB point of this probe
<azonenberg> it's past 6 GHz
<azonenberg> The pico probe is a little flatter than me out to 2 GHz but from DC to 500 MHz and 3.5 to 6 GHz ZI'm better
<azonenberg> and i suspect most of my not-flatness is the filter being a little too strong
<azonenberg> if i just tone down the peak a bit and possibly shift the center freq up a little, i'll be very happy with this response
<azonenberg> I do have slightly more loading than the Pico probe but it's not bad at all (especially considering i'm a 10x probe not a 20x)))))))))))
<azonenberg> from about 1 to 5 GHz the loading on v1.3 is significantly less than v1.2
<azonenberg> One potential issue is that these improvements i've made at lower frequencies have increased the Q of the tip resonance
<azonenberg> each rev from 1.0 to 1.3 the resonance has gone up in frequency, been less wide, but been deeper
<azonenberg> Pink is the awful pico probe, blue is the good pico probe
<azonenberg> Cyan is the kickstarter rev, black was my last rev, red is the latest
<azonenberg> thie is zoomed in my probe S21 vs the Pico "6 GHz" probe. The big dip in mine out to 5 GHz is the response of the filter which seems to be too strong (It's supposed to be toning down peaking, not adding a dip)
<azonenberg> but i think calling my design a 6 GHz probe is completely reasonable at this point and further improvements will just be tweaking flatness
<azonenberg> (also ping wbraun, Kliment)
juli966 has quit [Quit: Nettalk6 - www.ntalk.de]
<azonenberg> and here's the return loss curve https://www.antikernel.net/temp/probe-v1p3-03.png
<azonenberg> red vs black (v1.3 vs v1.2) is better until past 5 GHz then the resonance hits
<azonenberg> S21 measured across 50 ohm terminator, S11 across open circuit
<lain> azonenberg: !! awesome
<azonenberg> i'm also very optimistic for response of the solder-in probes at this point because they don't have the tip needle which is the source of the remaining resonance, or the filter needed to counteract it
<azonenberg> I think those could very well end up being >6 GHz probes
<azonenberg> as in, i'll spec them as 6 GHz probes because i have no way to characterize them any higher :p
<azonenberg> lain: you can see just how atrocious the TA061 is compared to all the others lol
<azonenberg> massive notch in S11 and huge spike in S21 right in the middle of the passband
<azonenberg> like, i have a tip resonance but at least it's at >6 GHz and not in the middle of the operating range
<azonenberg> but yeah, my S21 from DC to 6 GHz is already flatter than the pico probe which had been my target to beat for quite a while
<azonenberg> And that's a $1K probe
<azonenberg> whatever i do after this is icing on the cake
<azonenberg> also hmmm, thinking more i don't think the dip at 2-3 GHz on mine is the filter
<azonenberg> the filter's peak is actually closer to 5 GHz
<azonenberg> So it might be something else. In any case this version is Good Enough (tm) for the short term
<azonenberg> lain: also 115ps 10-90, 79ps 20-80 *including cable effects* measured by the scope
<azonenberg> With the cable de-embedded, 74ps 20-80 through the probe. On a scope specced at 75ps 20-80
<lain> wow
<azonenberg> So basically i think i'm at a dead end with the design now, lol
<azonenberg> I literally cannot make it any better. Because if i do, i'll have no way to know it's better :p
<azonenberg> i need better instrumentation to refine it further
<azonenberg> i might spend a little while in sonnet trying to flatten out the response around 2 GHz but other than that, i'm calling it done
<azonenberg> The scope is specced at 100ps 10-90 but for some reason when i de-embed the cable in glscopeclient i get strange results for 10-90 rise time so i can't quote a 10-90 through the probe
<azonenberg> i need to investigate more
<azonenberg> but 115ps 10-90 on a scope with 100ps datasheet spec before de-embedding the cable is already really good
Degi has quit [Ping timeout: 260 seconds]
Degi has joined #scopehal
Pretzel4Life has joined #scopehal
Pretzel4Ever has quit [Ping timeout: 246 seconds]
electronic_eel_ has joined #scopehal
electronic_eel has quit [Ping timeout: 260 seconds]
_whitelogger has joined #scopehal
<Kliment> azonenberg: how are the solderable probes meant to be used
<azonenberg> Kliment: It's a 150 mm (for this prototype, actual length may change) flex PCB with 1mm spaced signal and ground contacts at the tip, and a SMA at the other side
<azonenberg> and a 3d printed TPU strain relief/protective sleeve over most but not all of the length of the flex
<Kliment> That seems quite long
<azonenberg> The LeCroy WaveLink solder-in tips are 9 inches
<azonenberg> i decided to try a little shorter to start
<Kliment> is u.fl too lossy for this application?
<azonenberg> it would be easy to make several versions, the design scales easily
<azonenberg> You're getting ahead of me, hold on :)
<azonenberg> Anyway, the intention is for you to secure the tip of the probe and the very end of the TPU sleeve to the DUT with kapton tape
<azonenberg> then solder the shortest possible jumper from the tip to the test point and from the ground to the nearest ground
<azonenberg> using 30ga wire, the circuitmedic square wire i like, etc
<Kliment> yeah, I get how it works on the PCB end
<azonenberg> i may also make a version with castellations on the tip, there are pros and cons. my previous version used them and i want to try without
<Kliment> (though I'd not use it that way)
<azonenberg> anyway, that's the AKL-PT2
<azonenberg> The AKL-PT3 is a tiny probe about the size of a dime
<azonenberg> with the same solderable contact area as the PT2, the same attenuator string, and a U.FL
<Kliment> What I would do is solder it in directly - apply paste to traces, and heat it through the flex with an iron
<azonenberg> Kliment: you mean upside down? that works if tip and ground are exactly 1mm apart
<azonenberg> using it like an LGA
<azonenberg> if you had a suitable test point then great
<azonenberg> But if i intended it to be used that way i'd put a castellation on the end so you could have a solder fillet to get better contact
<azonenberg> Long term, i will probably offer both variants
<azonenberg> the PT2 prototype is intended to test the concept of top side contacts only
<azonenberg> as well as be a mechanical concept for the idea of a TPU-sleeved flex board as a probe
<Kliment> I don't see why you can't support both
<azonenberg> The point of the top side contact is you can put it on top of components or pads
<azonenberg> and not short
<azonenberg> also, less capacitive loading than castellations
<Kliment> yeah, but it's flex, you can just flip it around
<Kliment> and you don't need castellations to solder contacts through flex
<azonenberg> well the reason flipping it might be hard is that there are ~500um high attenuator resistors *right* next to that contact
<Kliment> Ah
<azonenberg> you'd need a very sharp bend in the flex and there would be lots of shear force on the joint
<Kliment> yeah that will be a problem
<azonenberg> minimizing distance from DUT to resistor string has significant impacts on frequency response
<Kliment> I guess via is not an option?
<azonenberg> there's a resonance in the AKL-PT1 i've been slowly pushing up and up in frequency by shortening the tip
<azonenberg> now it's around 6 GHz
<Kliment> via through flex will be like 0.12mm
<azonenberg> If i had a full via, i might as well cut the board there and have a castellation
<azonenberg> Which is what the very first prototypes i did back in 2017 had
<Kliment> cutting the board makes it a pain to solder
<azonenberg> why?
<Kliment> because there's no flat bit to press on
<azonenberg> you tape it down then just solder to the edge of the via like a chip resistor or something
<azonenberg> all you have to do is get a tiny bit of solder on the PCB trace and then heat the via a tiny bit, it wicks right up
<azonenberg> i've done it and it works great
<Kliment> I'd honestly rather have a full via and a trace-shaped pad underneath
<azonenberg> the other thing you can do is flux the area, tin the iron, then just tap the tip of the iron right at the castellation
<azonenberg> and boom it's soldered
<azonenberg> anyway AKL-PT2 v0.1 - although it wasn't called that yet - was made at pcbway and had about 30% yield due to the small via on the castellations requiring about +/- 100um tolerance on the edge cuts
<azonenberg> v0.2 was made at multech and had very good yield due to a combination of using a nicer fab and enlarging the vias slightly
<azonenberg> v0.3 is being done at oshpark and is top contacts only
<azonenberg> like i said I'm PoC-ing a lot of different design concepts in this one board
<Kliment> makes sense
<azonenberg> I do want to trial the via-and-bottom-contact approach
<Kliment> I think it will have better contact and better mechanical stability
<azonenberg> v0.1 and 0.2 were both in 2017
<Kliment> with the bottom contacts
<azonenberg> So the other reason is, via size
<azonenberg> at oshpark at least you need ~550um vias
<azonenberg> how do you propose to get a probe on a single pin of a 500um pitch tqfp with that?
<azonenberg> your tip contact has to be 250um or so in size
<azonenberg> and you need a way to get a ground contact which could be at any unpredictable location wrt the signal if you're debugging and not probing a by-design test point
<azonenberg> One of the things i want to do with the PT2 and PT3 prototypes is to experiment with probing random signals on real boards
<azonenberg> and actually see how the ergonomics work if i just pick say the RGMII TX_D3 signal on a random devkit
<azonenberg> and say i want to probe it
<azonenberg> Right now prototypes of the PT2 and PT3 are at oshpark expected to be back from fabMonday
<azonenberg> So based on test results we'll see how it goes. These are also my first designs on oshpark flex so my impedance calculations etc might also need some work
<azonenberg> The other two probes in the pipeline, right now just at the concept stage, are the AKL-PT4 and AKL-AD1
<azonenberg> the PT4 is basically the tip of a PT1 with a MMCX connector and a... M4 I think? screw stud for attaching to a sensepeek pcbite probe holder
<azonenberg> and the AD1 is an active differential probe, details TBD
<azonenberg> my preliminary concept for the AD1 is similar to various other active probe architectures where you have a solder-in tip containing a damping resistor and then an amplifier module that plugs into it
<azonenberg> I think the LMH3401 might be a good starting point for it
<azonenberg> the mating connection from the tip to the amplifier is TBD, and the proposed tip design is similar to the PT2 except with a differential pair instead of a coplanar waveguide
<azonenberg> not sure if i'd have just a bottom side ground or anything on top too
<azonenberg> but it would be a tightly edge coupled diffpair
<azonenberg> I haven't done a lot of design work on it yet, the plan was to spend some time studying the lecroy probes and thinking about what i did and didn't like about them
<azonenberg> And also get the passive probes refined to the point i was happy with them before branching out
<Kliment> azonenberg: you can have your vias further apart than your pads
<Kliment> azonenberg: be aware with the oshpark flex that they don't use coverlay like everyone else
<Kliment> azonenberg: they use LPI soldermask which has different electrical properties and is much less abuse-resistant
<azonenberg> Kliment: My prototype is maskless for that reason
<azonenberg> and i'll just accept the ENIG losses which, on a 6 inch line, will be considerable
<azonenberg> again it's a prototype
<azonenberg> i expect significant changes before it's final
<Kliment> maskless is unpleasant on a probe
<azonenberg> that's part of what the TPU shell is for
<azonenberg> so its less likely to short against stuff
<Kliment> Yeah
<azonenberg> but it has opens between the strain relief elements
<azonenberg> The v0.3 prototype as it stands is not intended to be something i could just start deploying as is. Think of it like a concept car
<azonenberg> The technology it demonstrates will eventually make it into the final probe
<azonenberg> but it may look nothing like it
<azonenberg> Key goals for this demonstrator are... Test the ergonomic aspect of top-only contacts on a solder-in probe
<azonenberg> validate mechanical properties of oshpark flex for this application
<azonenberg> Test the fit and snugness of 3D printed TPU as a strain relief and anti-short guard for flex PCB
<azonenberg> Figure out a way to DIY FR4 stiffeners for oshpark flex so i can use edge launch connectors with them for prototyping
<azonenberg> Get a feel for the mechanical properties of the LPI soldermask on the bottom side ground plane
<azonenberg> and yes, see what kind of bandwidth is achievable on the prototype as is
Nero_ has joined #scopehal
Nero_ is now known as NeroTHz
<NeroTHz> monin
<azonenberg> o/ NeroTHz
<azonenberg> Great test results from today :D
<azonenberg> v1.3 probe prototype in red vs PicoConnect 921 in blue, as well as a couple of other previous hardware revs and the Pico TA061 piece of junk
<azonenberg> It's got a little dip around 2 GHz i'd like to eliminate eventually but it's better than -3 dB to 6 GHz which is the upper end of my VNA range :D
<azonenberg> and that dip is nothing you can't de-embed, perfect flatness is impossible anyway
<azonenberg> 74ps 20-80% rise time after de-embedding the minicircuits cable when probing the LeoBodnar pulse generator
<azonenberg> On a scope specced at 75ps rise time
<NeroTHz> one second
<azonenberg> NeroTHz: that was a very long second
<azonenberg> :P
<NeroTHz> oh
<NeroTHz> right
<NeroTHz> yeah
<NeroTHz> :p sorry, work and distractions n shit
<NeroTHz> but the performance seems nice. I agree, don´t spend too much time on that 2GHz dip
<NeroTHz> rip
<NeroTHz> my work machine that I VNC into just stopped responding
<NeroTHz> seems like i´m gonna have to jump to a higher priority on that new-computer list at work
<azonenberg> NeroTHz: lol fuun
<azonenberg> i had to replaec a work machine a year or so ago due to random freezes too. some pcie device was randomly disconnecting
<azonenberg> that was soldered to the motherboard
<azonenberg> so the link ber was probably going bad due to a serdes failure or something
<azonenberg> but yeah i think at this point mission accomplished on the handheld probe. I can't make it any better without higher b/w test equipment
<azonenberg> So current plan is to send this prototype to shahriar and have him do a video on the project, order a batch of boards soon (budget is a bit tight thanks to construction on the house so a production-sized run might have to wait a little while)
<azonenberg> and shift R&D efforts to the solder-in probes as well as probably an active differential probe at some point
<NeroTHz> ah that will be interesting to see
<azonenberg> The solder in probes are mostly a mechanical design problem
<azonenberg> i can reuse my existing well tested attenuator
<azonenberg> and the tip resonance will be way high out of band because of the much shorter "tip" which is just a solder pad
<azonenberg> For the time being though, this probe is better than my test equipment. So i think i'm gonna make a couple more of them and actually start using them for projects
<azonenberg> (right now i have 2 bare PCBs plus the one I'm sending to TSP
<azonenberg> the one other little thing i might want to try is seeing if i can source where Pico gets their probe tips for the 900 series from
<azonenberg> those tips are slightly skinnier than mine so i could use a smaller socket with less capacitance
<azonenberg> and finer wire for solder-in applications
electronic_eel_ is now known as electronic_eel
<NeroTHz> ffs
<NeroTHz> people who first do a simulation of a PA, publish it
<NeroTHz> then a year later fabricate it, then just re-write the paper (edit your sentences here and there so google/plagerism software wont catch it) and republish it
<Degi> PA = power amplifier?
juli966 has joined #scopehal
<azonenberg> Totally unrelated, I just found an error in the wavelink low-bandwidth probe datasheet (D4xx/D6xx)
<azonenberg> on page 7 there's a contradiction as to the performance of the Dx10/Dx20-QC probe tip
<azonenberg> the rise time numbers for the D410/D420 list SI, QL-SI, HiTemp, QC, and PT on the first line then QC again on the second line
<azonenberg> based on QC being listed by itself there, and again for the D610/D620 amplifier modules at left, it appears that this is the correct number and its inclusion on the first row is an error
<azonenberg> Already reported to them so they can fix it
<azonenberg> Flex probe prototypes just got received by oshpark
<azonenberg> hopefully should have them early next week
bvernoux has joined #scopehal
bvernoux has quit [Quit: Leaving]
<azonenberg> So this is interesting
<azonenberg> reading through more lecroy documentation
<azonenberg> "WaveLink Medium Bandwidth probe amplifiers have no gain or attenuation control. However the WL-PLINK platform/cable assembly does provide gain and attenuation controls"
<azonenberg> so the wavelink adapter may not be 100% passive
<azonenberg> unsure if the ProBus version is, or if there's some intelligence there too
<azonenberg> Also, it looks like ProLink goes up to about 20 GHz and then they switch to 2.92mm connectors
<azonenberg> WaveLink seems to have two variants, low bandwidth is 4/6 GHz and medium bandwidth is 8/10/13 GHz. There does not appear to be a "high bandwidth" wavelink offering
<azonenberg> instead you move to the DH series
<azonenberg> which is a separate probe system that has 8 GHz available with probus2/prolink, then 13, 16, 20 GHz ProLink and 25, 30 with 2.92 mm
<azonenberg> The DH probe amps are significantly larger than WaveLink ones and have a bunch of buttons on them, vs the WaveLink ones which have no physical controls
<azonenberg> also interesting... WaveLink handheld browser tips go to the full bandwidth of the amp. But DH browser tips max out at 16 GHz, while the DH solder-in tips go to 30 GHz
<azonenberg> I'm impressed they got 16 GHz out of a handheld browser probe considering the difficulties I had with tip resonances on mine
<azonenberg> There are mechanical constraints that make significantly smaller tips virtually impossible, so I wonder if they're either using lossy tips to damp out the resonance or doing some kind of compensation on the probe side of the tip to damp it
<azonenberg> also interesting, WaveLink tips appear to be entirely passive and you need to tell the scope which one you are using to get correct gain display etc
<azonenberg> But DH tips contain some kind of identification mechanism so the probe/scope knows which one you are using
<azonenberg> Wonder why they didn't just slap a 24C on the WaveLink tips?
<azonenberg> they could power it up, read the idcode, then power it down and have zero contribution to noise
<azonenberg> and it's not like a 24c will increase the cost of a $8K probe significantly