azonenberg changed the topic of #scopehal to: libscopehal, libscopeprotocols, and glscopeclient development and testing |,, | Logs:
<azonenberg> given that i verified the GCPW geometry and SMA launch already i'm suspecting it's the tip that's reflecting mostly
<azonenberg> So i'm running additional sims of that
_whitelogger has joined #scopehal
<azonenberg> ok, so it looks like the probe tip isn't super well matched. The geometry i have right now hits 81 ohm impedance @ 2.38 GHz if extended out to a longer box for my testing
<azonenberg> Time to re-optimize that then i guess
Degi has quit [Ping timeout: 246 seconds]
Degi has joined #scopehal
_whitelogger has joined #scopehal
juli965 has quit [Quit: Nettalk6 -]
<azonenberg> still running parameter sweeps but it looks like the current tip socket is definitely not well matched on the oshpark stackup
_whitelogger has joined #scopehal
Famine- has joined #scopehal
<Famine-> azonenberg, do you know if 200pS pulse gens are still ITAR restricted ?
smkz has quit [Ping timeout: 256 seconds]
LeoBodnar has quit [*.net *.split]
anuejn_ has quit [*.net *.split]
electronic_eel has quit [*.net *.split]
_whitelogger has joined #scopehal
bvernoux has joined #scopehal
<bvernoux> hello
<bvernoux> I have found a very interesting RF microwave Synthesizer
<bvernoux> Specifications are very nice
<bvernoux> The chipset cost is a bit expensive about 80USD ... but anyway to go from 172Mhz to 18GHz with such specifications it is hard to find better for that price
<Degi> Where can you get that chip for qty 1?
<electronic_eel> bvernoux: what are you planning to do with it?
<bvernoux> good question ;)
<bvernoux> some microwave experiments ;)
<Degi> Hmh how about a SDR
<Degi> With like 500 MHz baseband or so
<bvernoux> but I'm searching since lot of time a good wideband synthesizer up to 18GHz or more ;)
<bvernoux> and this one have amazing spec
<Degi> A good IQ mixer which doent cost like a k would be neat for that
<bvernoux> yes a good mixer is required too
<bvernoux> I would like to build a good up/down converter with such synthezier
<Degi> Only up to 6 GHz tho
<Degi> Yes
<Degi> The IQ modulator seems to be kinda hard
<bvernoux> what is nice in IDT RF Synthesizer is the fact you can output up to 10dB ;)
<bvernoux> up to 18GHz
<Degi> Like if you IQ modulate your baseband -250 to 250 MHz to 5-5.5 GHz and then you have a mixer which can convert that to 0-18 GHz but then you have image problems
<Degi> Neat
<Degi> Maybe add a TWT amplifier
<bvernoux> yes it needs some filters ...
<electronic_eel> hmm, if you look how much time it takes azonenberg to get something "simple" like just a sma footprint to work well to something like 6 GHz, I think getting something more complex like a whole synth to work well up to 18 will take much more time and money (=custom rogers pcbs)
<Degi> Like you'd need a frequency adjustable filter for the image
<bvernoux> I plan such synth to use other connectors ;)
<Degi> Eh, azonenberg wants to make a very good probe which is kinda long too
<bvernoux> like my southwest 3.5mm ;)
<bvernoux> which are specified up to 40GHz ;)
<Degi> Like especially if your signal is binary data, you can get away with a lot, like FR4 to 4+ GHz, bends all around and vias everywhere
<bvernoux> without surprise of crappy SMA connectors which never reach more than 6GHz ;)
<bvernoux> yes for such design it clearly needs RO4350B ...
<bvernoux> it is a good test for my future SA which shall arrive in about 1 month
<Degi> Unless you put the chips really near to eachother
<Degi> Is it the one that can do like 20 GHz?
<electronic_eel> but do you think making the nice 3.5mm connectors working well to 18 will be easy? I think aligning the footprint will still be difficult
<bvernoux> when you see the price of components I will do not take risk to use a cheap FR4 PCB ;p
<bvernoux> Ro4350B are quite affordable today
<bvernoux> and it's definitely not a cheap board
<Degi> Ookay lol
* Degi put away 2 € PCB
<electronic_eel> 2 € pcb = 2 GHz, 18 € pcb = 18 GHz ;)
<bvernoux> but yes in theory with very short length FR4 could do the trick ;)
<bvernoux> it is just I will be afraid too loose a lot of insertion loss at connection points because of FR4 and ENIG ...
<electronic_eel> bvernoux: about something completely different - did you have time yet to dissect the logic analyzer input circuit?
<bvernoux> when you see the effect up to 6GHz you imagine the crap up to 18GHz ;)
<bvernoux> electronic_eel, ha yes good things I need to open one
<Degi> Well of course you would leave the silkscreen on xD
<Degi> Or use organic solder preservatives
<bvernoux> but to do good measurement I need a special PCB to test it with my VNA ;)
<bvernoux> to retrieve full caracteristics
<electronic_eel> you mean a test jig pcb, to compare the results with la probe and without?
<bvernoux> yes
<bvernoux> a bit like azonenberg PCB test fixture for probe
<electronic_eel> yes, that would be really handy for developing the CONWAY input circutry as well
<bvernoux> I think of a basic test
<bvernoux> as anyway such probe shall not exceed 200MHz ;)
<bvernoux> just to plug each side on a basic test fixture
<bvernoux> with sma on both side
<Degi> TIL its surprisingly hard to blow up your PCB by sending random data over PCIe
<bvernoux> directly on VNA ;)
<bvernoux> before to explode the cable to check the resistor/capacitor they have put
<bvernoux> I suspect it is just a little res in fact
<electronic_eel> hmm, couldn't azonenbergs probe test pcb be reused for this? I know it didn't perform as well as he planned for the scope probe, but maybe it can still be reused for the lower-speed la
<bvernoux> yes it could maybe
<bvernoux> need to build one with my new connectors southwest ;)
<electronic_eel> I think I remember that he said that he had still some unpopulated pcbs left of it
<bvernoux> I have not found footprint for southwest connector for KiCad need to design it
<bvernoux> if someone has I will gain time ;)
<Degi> Urks wtf is up with this mainboard, why do different pcie slots behave differently
<bvernoux> anyway they are not too complex
<electronic_eel> Degi: welcome to crappy bios / mainboard world
<bvernoux> but I will like to also design perfect match from transition to PCB
<bvernoux> so it could be part of the southwest footprint
<Degi> Like the PRSNT signal is wired wrongly or so on the x16 slot and the rx CDR outputs a different frequency than it does on the x1 slot and aaah
<electronic_eel> Degi: some pcie slots are often directly connected to the cpu, the others to the chipset, so that is likely where the clock difference comes from
<Degi> THe thing is they should be the same heh
<Degi> Maybe ill just stick it into my PC and hope it doesnt damage anything
<bvernoux> just a thing I do not know is my southwest connectors are 1092-0xA-5
<bvernoux> but i do not know exactly which one ;)
<Degi> Wow, it even gives partially meaningful symbol sets... Maybe I just need a new mainboard
<bvernoux> x=> can be 1, 2, 3 or 4
<bvernoux> it is mainly some variance on pin size
<electronic_eel> bvernoux: didn't the shop where you bought them spec that?
<bvernoux> no
<bvernoux> and there is no ref on the connectors
<bvernoux> except SOUTHWEST which clearly seems genuine
<electronic_eel> did you ask them? often you can get scans of the manufacturer bag label or stuff like that
<bvernoux> and they are quite heavy too
<bvernoux> only the launch pin differ in fact
<bvernoux> from .009 up to .02 inch ;)
<bvernoux> for outer size
<bvernoux> maybe with a good microscope and high resolution I could measure that ;)
<bvernoux> the steps are 0.009, 0.012, 0.015 and 0.020 ;)
<electronic_eel> maybe put it on a diffraction grating pattern or something fine with a known precise pattern and then photograph it through a macro lens or microscope
<bvernoux> yes anyway it is tricky
<bvernoux> at least I can see the extreme ;)
<bvernoux> 0.009 vs 0.02 ;)
<bvernoux> I doubt the guy which sell them know the exact ref anyway
<bvernoux> it is a surplussellet
<bvernoux> they do not know what they sell ...
<bvernoux> and it is possible that I have received mixed version ;)
<bvernoux> it is a shame that Southwest does not have a ref on each connector for that
<electronic_eel> ok, now I understand why azonenberg refered to your source as a shady seller and did not consider them
<bvernoux> anyway for the price I cannot say anything ;)
<bvernoux> I will caracterize myself the connectors
<bvernoux> to be sure they are good and for 4USD per connector in all case it is perfect
<bvernoux> when you know such 2.92mm Southwest microwave connectors cost 120USD/unit
<bvernoux> also after inspection they are brand new and never used
<bvernoux> and they shall be even better than my genuine 142-0771-831 SMA High Freq End Launch Johnson (Cinch Connectivity) BelFuse ;)
<bvernoux> especially easier to connect with better performances
<bvernoux> ;)
<bvernoux> let's buy 10 mores for 45USD ;)
<bvernoux> bought ;)
<bvernoux> Total 59.28USD with shipping for 10 more Southwest Microwave 2.92mm ;)
<bvernoux> it is the price of SMA connector not even reaching 6GHz on DigiKey.... ;)
juli965 has joined #scopehal
juli965 has quit [Quit: Nettalk6 -]
<bvernoux> electronic_eel, answer from surplusseller => I have no idea which is ... ;)
<bvernoux> electronic_eel, the only way is to measure the pin ;)
<electronic_eel> wow, they were fast with their answer, unfortunately it wasn't helpful
<electronic_eel> so, how much does this length matter?
<miek> do you have enough to sacrifice one? :p
<miek> could cut the pin out and measure it with a micrometer
<miek> or, does it stick out enough anyway? hard to tell from this drawing
<electronic_eel> hmm, I don't find any 1092-0xA-5, just 1092-0xA-6, -7, -8
<azonenberg> Famine-: you can get LeoBodnar's 40ps pulse gen from the UK and sidestep ITAR entirely. had no trouble buying it
<azonenberg> no idea if one made in the US would be controlled
<bvernoux> I have finally received my JFW 50DR-068 ;)
<bvernoux> but when turning knob it is like it go to far without doing the "click"
<bvernoux> like something move inside ...
<bvernoux> will tighten the screws ...
<miek> is the shaft definitely rotating? it could just be that the set screw in the knob is loose
<bvernoux> yes after check it is because screw are not tighten enough
<bvernoux> it's a fat cylinder ;)
<bvernoux> I prefer my other one
<miek> ah good :) much better problem to have than something inside being wrong..
<bvernoux> I definitely prefer the Model 50BT-001
<bvernoux> with a box
<bvernoux> 50BR-001
<bvernoux> 50DR-068 is intended to go to 2.4GHz or more
<bvernoux> so I will check
<bvernoux> it is a custom version not available on official website
<bvernoux> it is dual rotor with 70dB
<bvernoux> it comes from Israel ;)
<bvernoux> hehe nice I have found a Torx T5 and I have aligned everything ;)
<bvernoux> let's check it with the VNA now
<azonenberg> LOL
<azonenberg> So, pink line is what the latest prototype uses WRT tip geometry
<azonenberg> red line is what i plan to use on the next pcb rev
<azonenberg> look a little bit better?
<azonenberg> here's S21. This is for the tip itself extruded out to a 20mm long box so i can get a better measurement off it
<azonenberg> and equivalent impedance
<azonenberg> not sure how the tip didn't get tuned before
<azonenberg> i swear i did
<azonenberg> anyway, so i'm going to redo the probe and send out v0.9 with this tip geometry and see how that works...
<azonenberg> current tip is 0.2mm gap and a 2.1mm wide ground plane cutout under it
<azonenberg> new is 0.125mm gap and 1.5mm cutout
<bvernoux> azonenberg, red line is clearly better ;)
<bvernoux> a huge improvement
<azonenberg> Yeah. I still have to optimize the transition from the fat line to the CPW
<azonenberg> but this is progress
<bvernoux> you speak about probe-tip-match2.png
<bvernoux> which is not flat from 1.25Ghz to 2.7GHz
<bvernoux> there is clearly something wrong between those 2 freq
<bvernoux> it is quite nice from 3Ghz to 6Ghz
<bvernoux> even if it reach -1dB
<bvernoux> I imagine it is with ENIG OSHPARK PCB ?
<bvernoux> ha yes it is wrote ;)
<azonenberg> this is all simulation with bare copper on oshpark fr408 stackup
<azonenberg> i'm not modeling the enig effects
<bvernoux> yes anyway the effect of OSHPark 4 Layers are something like 2dB for 1 inch
<azonenberg> I need to fine tune more. Rolloff is fine, weird notches are not
<bvernoux> when in simulation you have something like 0.6dB IIRC
<bvernoux> the impedance is quite nice
<bvernoux> but we clearly see there is something betwen 1.25GHz to 2.7GHz
<bvernoux> which clearly show it
<azonenberg> I'm not done optimizing
<bvernoux> could be the matching on the SMA ...
<bvernoux> anyway the good things is improvement are huge vs old blue curve
<azonenberg> No, this is simulation of a thru line with no connectors
<azonenberg> just extruding the tip connector out to be about 20mm long
<azonenberg> so i can get a full wavelength or so of length to model
<bvernoux> ha ok
<azonenberg> Here's VSWR. Still a spike around 2 GHz. I'm running more sweeps with tighter CPW gaps and smaller cutouts
<bvernoux> just 0.1dB
<bvernoux> swing
<bvernoux> yes there is something with 2GHz
<azonenberg> this is actual vna data of a ~40mm thru line on the same CPW stackup as the probe with the new amphenol sma at each end
<azonenberg> not perfect but halfway decent, I think i can probably still optimize the match a little bit though
<azonenberg> but oshpark's stackup not being controlled impedance makes it hard
<azonenberg> i think if i switch to 4350B with impedance control this geometry will do a lot better
<bvernoux> yes I doubt you can do something better with OSHPark 4 Layers ENIG imperfections especially for that
<bvernoux> yes you will have something more flat
<bvernoux> for a high end version ;)
<bvernoux> 0.2dB is not so big
<bvernoux> especially on full scale from 0 to 6GHz
<azonenberg> that last one was VSWR not S11
<bvernoux> ha ok
<bvernoux> i was thinking it is S21 ;)
<azonenberg> max S11 at 5.1 GHz is -19 dB
<bvernoux> so it is different ;)
<bvernoux> -19dB S11 is nice
<bvernoux> especially for worst case
<azonenberg> S21 on the thru line is a nice flat rolloff from 0 at DC to -1.7 dB @ 6 GHz
<azonenberg> so i'm pretty confident the sma and cpw geometry are now well matched to the oshpark stackup
<azonenberg> now i just have to fine tune the attenuator and tip
<bvernoux> great
<bvernoux> I prefer to check S21/S11 than VSWR ;)
<bvernoux> VSWR is more for HAM guys
<azonenberg> well i linked s11/s21 earlier
<azonenberg> and yeah i normally work with s-params too
<bvernoux> ha ok it corresponds to
<bvernoux> Z=50Ohms
<bvernoux> then it go to about 55Ohms at 2.5GHz
<bvernoux> it is linked to tip/attenuator
<azonenberg> This simulation is of the tip only, extruded out to be 20mm long
<azonenberg> it's not the full probe
<bvernoux> 404 Not Found
<azonenberg> try again
<bvernoux> ha yes we see exactly the same
<azonenberg> This is zoomed in on the graph for the current geometry, you can see there is a higher impedance at higher freq. I think this means the ground plane cutout is still a little too large so i'm running another sweep with smaller
<bvernoux> when it exceed 2GHz there is a mismatch ...
<bvernoux> if you can flatten that it will be perfect
<monochroma> come on azonenberg, flatten the curve! ;)
<azonenberg> lol
<bvernoux> hehe
<bvernoux> and you could advertize it to be a 6GHz probe ;)
<bvernoux> and maybe even more ;)
<bvernoux> but the hard part is to find a VNA which is > 6GHz ;)
<azonenberg> First step is going to be optimizing this, then making another probe spin and waiting another ~2 weeks
<azonenberg> i also am adding a second ground socket at the tip
<bvernoux> you are always using OSHPark 4 Layers ?
<azonenberg> For these tests yes
<azonenberg> i have two older spins on multech 4350B
<bvernoux> yes interesting anyway that just impact the loss
<azonenberg> but it's expensive and slow so i'm holding off on that until i've fine tuned the design more. I thought I was closer than I was
<azonenberg> exactly
<azonenberg> more importantly 4350b and 408hr have almost the same Er
<bvernoux> and probably add some "swing" but not too much
<azonenberg> so it should be basically a drop-in substitute
<azonenberg> it's a difference of like 0.02
<bvernoux> yes it is great for that
<bvernoux> for RO4350B you are planning to use Immersion Silver ?
<azonenberg> Yes
<bvernoux> as It seems over time it is not perfect ...
<bvernoux> it seems it is why even on some RF microwave stuff we see ENIG or ENEPIG in fact
<bvernoux> ENEPIG is really better than ENIG
<azonenberg> does enepig still have the Ni+P loss issue though?
<bvernoux> I have quickly checked my JFW 50DR-068 ;)
<bvernoux> it is not bad and work fine up to 2.4GHz with error of max 1dB ;)
<azonenberg> yeah but when you're making a probe that you want flat <1 dB to several GHz...
<bvernoux> from attenuation 0dB up to 70dB
<bvernoux> azonenberg, it seems ENPIG is not affected by Ni+P loss
<azonenberg> do you have citations for that?
<bvernoux> let me check the things I have found on it
<azonenberg> because the palladium layer is really thin
<bvernoux> with some curve up to 40GHz ;)
<azonenberg> i would not expect it to pull much signal out of the nickel
<azonenberg> oh, thoughts on the rigol dsa875?
<bvernoux> here
<azonenberg> i see it appearing in a lot of signalpath videos, sharihar seems to like it
<bvernoux> so ENEPIG is better than ENIG ;)
<bvernoux> but Immersion Silver is clearly better ;)
<azonenberg> i normally have not been a fan of rigol gear though
<azonenberg> bvernoux: got a timestamp for me in that vid?
<azonenberg> i've watched the whole thing but i dont remember enepig being mentioned
<azonenberg> oh here we go, 23:01
<bvernoux> 24
<bvernoux> 24:18
<bvernoux> is good
<bvernoux> we see all materials
<azonenberg> so yeah enepig is slightly better than enig, as expected from the palladium layer above the nickel putting more signal in the outer plating - 0.2 dB @ 40 GHz better
<bvernoux> ENIG/ENEPIG are very similar up to 6GHz anyway
<azonenberg> but immersion silver, bare copper, and OSP are basically indistinguishable
<bvernoux> yes it is the good point of this video
<bvernoux> Immersion Silver shall be better than Copper too in theory
<bvernoux> but it is not in reality ;)
<azonenberg> well its super thin and only a little bit more conductive, so while it's better the difference is lost in the noise floor
<bvernoux> yes they explain that in video
<azonenberg> yeah i've watched the whole thing. and read the conference paper
<bvernoux> the fun is SolderMask is less bad than ENIG/ENEPIG ;)
<azonenberg> anyway, thoughts on the dsa875? not planning to get a specan any time soon
<azonenberg> but it seems surprisingly good for something from rigol
<bvernoux> I would not recommend DSA ;)
<bvernoux> they are limited
<azonenberg> oh?
<bvernoux> BW ...
<azonenberg> that one goes to 7.5 GHz
<bvernoux> they have a very good DANL anyway
<bvernoux> but BW is quite limited vs other SA
<bvernoux> and nothing to stream in realtime
<bvernoux> so definitely not a RTSA
<azonenberg> They say "PC software available" but i havent looked for a scpi programmers manual
<azonenberg> well yeah
<bvernoux> they sw are bad ;)
<bvernoux> their
<bvernoux> usually they are slow ...
<azonenberg> oh of course. whatever specan i get i'll want to plug into scopehal
<bvernoux> I will recommend the KC908 ;)
<bvernoux> I'm waiting mine Batch 2
<azonenberg> sharihar seems to like those too, but do they have pc interface capability?
<bvernoux> up to 22GHz dual chan which can do localization between 2 channels
<bvernoux> computing phase ...
<bvernoux> very nice it is not common ;)
<bvernoux> KC908 is their 1st SA
<bvernoux> a special SA
<bvernoux> it is fully supported on GRC with USB3.0
<azonenberg> no ethernet though?
<bvernoux> let me check
<bvernoux> IIRC 1Gb ETH
<bvernoux> it seems there is no ETH
<bvernoux> but 15MHz BW IQ streaming
<bvernoux> which can be unlocked to 30MHz over USB 3.0
<bvernoux> with degraded performances
<bvernoux> as the filter are not as good with 30MHz BW
<bvernoux> The spec is clearly not complete =>
<bvernoux> Batch2 contains a new RF PCB
<bvernoux> optimized
<bvernoux> mixed with RO4350B ...
<bvernoux> the fun is it is portable ;)
<bvernoux> The guy developing Deepace stuff have lot of interesting news here
<bvernoux> I translate from chinese to english ;)
<bvernoux> it is like a blog
<bvernoux> here we see it is fully supported by HDSDR
<miek> it still feels like a bit of a gamble to me. when i'm paying real money for test equipment, i want a bit more confidence in the specs than they're giving
<bvernoux> the fun is
<bvernoux> HDSDR connection 908, with the best computer, can barely work to 20M bandwidth. But connecting to GNURADIO can easily work up to 40M bandwidth. It is generally recommended that HDSDR be set to 5M bandwidth.
<bvernoux> so it can even stream over USB 3.0 at 40MSPS ;)
<bvernoux> 40MHz BW ;)
<bvernoux> so 80MSPS ;)
<bvernoux> miek, yes it is why it is like a "crowdfunding"
<bvernoux> the spec are not frozen and not full
<bvernoux> but for that the price is 1513USD instead of about 4KUSD for final version
<bvernoux> for 1500USD there is absolutely nothing on market to compete with such things
<bvernoux> especially it is fully portable and autonomous ;)
<bvernoux> and Batch2 unlocked to 22GHz ;)
<bvernoux> I'm pretty sure it will be better than those old Agilent stuff E440x ;)
<bvernoux> Especially for fast display and which can be used like a SDR so it is clearly more than just a simple SA
<bvernoux> miek, but yes I understand there is a risk ;)
<bvernoux> miek, so far I was quite disappointed by the old instruments we find on Ebay as they are loud, heavy and SW are so old that it is a pain to interface with a PC to retrieve data ...
<bvernoux> and the fact that they can break anytime without any warranty ;)
<azonenberg> bvernoux: this is why i buy official refurbs
<azonenberg> cheaper than new, more reliable and better support than random ebay stuff
<azonenberg> seems like a good middle ground
<bvernoux> like it was the case for my Agilent E4405B which came broken
<bvernoux> azonenberg, yes official refurbs are even better but quite expensive
<azonenberg> well as i see it, the three options are new, official refurb, or third party used seller
<bvernoux> and so far limited to scope with Lecroy ;)
<azonenberg> other vendors don't have similar deals?
<azonenberg> TRS-RenTelCo sells retired rental fleet equipment from any manufacturer with, i believe, a 1-year warranty
<bvernoux> Lecroy SA are just Rigol rebranded ;)
<azonenberg> (of their own)
<bvernoux> So I will do not trust Lecroy for SA
<azonenberg> So... you need to understand how lecroy works a bit
<bvernoux> They are just entering in the SA game
<azonenberg> Teledyne Test Tools is not real lecroy gear
<azonenberg> T3 is kinda like "linksys by cisco"
<azonenberg> just because your new wrt54g has a cisco label on it doesn't make it a real cisco router
<bvernoux> T3SA Series - Spectrum Analyzers look like Siglent SA ;)
<bvernoux> rebranded
<azonenberg> I know for a fact the wavesurfer 3000 series is rebranded siglent
<azonenberg> the waveace is i think an owon?
<azonenberg> wavejet is third party too. They outsource and rebrand all of their low end gear
<azonenberg> T3 anything is not real lecroy hardware
<bvernoux> ha yes so maybe it is the same for their SA
<azonenberg> If it says "teledyne lecroy" and not "teledyne test tools" it should be good
<bvernoux> hehe ;)
<bvernoux> so only Lecroy scope ;)
<bvernoux> and not the T3DS ;)
<bvernoux> IIRC their Voyager USB 3.x is not bad too
<azonenberg> Yes. T3DSO is the waveace/wavejet successor
<miek> a lot of the old ebay kit does come with service manuals though :) but yeah, if i were reliant on this gear for work i might go for different sources
<azonenberg> i believe it's siglent based
<bvernoux> miek, and on the other side we have chinese company doing RF like Deepace
<bvernoux> and they are clearly very good
<bvernoux> but not well known so far
<bvernoux> they started reversing other brand in fact
<bvernoux> but now they are doing their own stuff as they have learnt from the big company
<bvernoux> miek, you can check different "blog" of the CTO of Deepace here
<bvernoux> there is tons of interesting stuff very advanced with lot of details
<bvernoux> even if they are not doing open source stuff so far
<bvernoux> they are quite open in mind
<Famine-> azonenberg, sadly that doesn't have the output power i need
<bvernoux> miek, on the other hand the NanoVNA v2 seems really amazing for 50USD ;)
<miek> bvernoux: they do seem to be very good. my specific issue there is just that once i'm paying $1500+, it's because i really want to trust my measurements - so (for me) having things like an unbounded amplitude inaccuracy, or saying "we only specify one port, the other one could be different" isn't good enough (for me!)
<bvernoux> I plan to buy the NanoVNA 2 with 4inch screen for fast VNA measurements up to 3GHz ;)
<azonenberg> I'm still holding out hope that darrell's vna will be available this year
<azonenberg> i really want one of his
<bvernoux> of course that's still not comparable to HP VNA but can be useful when need fast measurements without too much accuracy ...
<bvernoux> azonenberg, ha yes me too
<bvernoux> azonenberg, especially the 20GHz 2chan version ;)
<Famine-> azonenberg, i found the most awful jankiest pulse generator ever on ebay and i really want it lol
<bvernoux> azonenberg, do you have news from Darrell ?
<azonenberg> no he's been off the radar for a while
<miek> bvernoux: yeah, it still is damn good for the price. if i had more need to look at stuff between 6-22GHz i'd probably have different opinions :)
<bvernoux> azonenberg, I see he's playing with 40gB ETH =>
<bvernoux> for the VNA ;)
<azonenberg> yeah i saw that
<azonenberg> but nothing re the VNA
<bvernoux> he promised their VNA will be quiet small and fast ;)
<bvernoux> and I hope with a good multiplatform GUI ;)
<bvernoux> with SCPI of course over Ethernet
<bvernoux> latest news was Dec 12, 2019
<bvernoux> At least a year until production, maybe even for public beta. Avoiding building too much hardware until things are finalized. A 2 port will be about 50 x 200 x 300 mm and 2 kg.
<bvernoux> it is for Darrell's VNA 2 Ports 20GHz version
<azonenberg> He talked to me a while back about me joining the beta program when he has units ready for that
<azonenberg> but hasnt sent me updates so clealry he isnt there yet
<bvernoux> yes he told he plan a Beta with rent
<bvernoux> I'm asking him by Email ;)
<azonenberg> ooook so, let's see... The software seems to be coming along nicely, anybody have short term priorities for me to work on with scopehal/glscopeclient?
<azonenberg> i know a bunch of folks are working on various new features, which is good, but you don't seem to need help at this time
<azonenberg> i plan to work on the RGMII protocol decode soonish, it's one of the older outstanding tickets
<azonenberg> the other one i need to do soonish is #1, allowing eye patterns on signals with a DC offset
<azonenberg> right now the eye decoder assumes everything is centered about zero. I actually built in some logic to center them, but never implemented the UI
<bvernoux> the USB HS decoder seems amazing too
<bvernoux> Do you have tested it ?
<azonenberg> yeah that's... miek i think working on it?
<azonenberg> i have not. I wrote the original FS/LS decoder and never had time to work on HS support
<bvernoux> yes it is a very good feature
<azonenberg> SS will likely be a separate decode as the PHY has nothing in common with FS/LS/SS
<azonenberg> FS/LS/HS*
<bvernoux> I'm checking Ebay auction for 4GHz Lecroy Scope ;)
<miek> yeah, still a bit more work to do before i can PR but it's looking good :)
<bvernoux> maybe something like TELEDYNE LECROY WAVERUNNER 640Zi 4GHz OSCILLOSCOPE 40 GS/s WITH PROBES
<azonenberg> Anyway, so i think that's where we stand on the software right now
<miek> i'll probably be able to test LS too, i've got some fairly flexible devices here that should be able to be forced to LS only
<bvernoux> Teledyne LeCroy HDO9304 3GHz 40GS/s 4Ch 10-Bit High Definition Oscilloscope will be even better but 20KUSD ....
<azonenberg> Hardware wise, I have a BLONDEL testbed on my left side lab bench that hasnt been touched in a few weeks
<azonenberg> so i want to get back to that
<azonenberg> then the MEAD board plus support stuff are at fab, stencils and components ordered
<azonenberg> Parameter sweeps running in sonnet now optimizing the tip geometry on the probe, hoping those will finish and i can order a respun probe today
<azonenberg> probably getting on the same panel as the MEAD stuff if i'm quick about it
<bvernoux> azonenberg, whats is the aime for MEAD board ?
<bvernoux> I do not see any description
<bvernoux> aime->aim
<azonenberg> i may not have pushed the updated readme yet?
<azonenberg> bvernoux: MEAD and CONWAY are interface-compatible logic analyzer pods. SFF-8087 mini-SAS connection to the scope providing 8 lanes of LVDS data, LVCMOS33 UART, and 12V power
<azonenberg> same comparators, 8-channel serial DAC for setting threshold level
<bvernoux> ha ok
<azonenberg> difference is MEAD is 50 ohm inputs and CONWAY will be high-Z
<azonenberg> MEAD takes inputs on MMCX connectors for transmission line probes
<bvernoux> so they are just intended to be used like La ?
<bvernoux> LA
<azonenberg> Yes. BLONDEL will have two ports for a total of 16 LA channels, as well as the 8 analog channels
<azonenberg> as will most of our future scopes
<azonenberg> MAXWELL is a parallel project, same 1U chassis as BLONDEL but it's a pure logic analyzer
<azonenberg> xc7k160t in ffg676, a sodimm of ddr3 1866, 40GbE, something like ten MEAD/CONWAY logic probe connections
<azonenberg> so ~80 digital input channels
<bvernoux> ha ok great
<bvernoux> for DDR4 ;)
<azonenberg> at 1.25 Gsps (MEAD/CONWAY are active probe heads without latching)
<azonenberg> but there will also be one more LA port on MAXWELL that samples at 10 Gsps on 4 channels
<azonenberg> using Kintex-7 GTXs
<azonenberg> DENNARD is a planned higher end LA that will use a much bigger part with more GTX and have 20+ 12.5 Gsps LA channels
<azonenberg> That will be suitable for DDR protocol decode etc
<azonenberg> MAXWELL is meant for lots of slow channels and a very few fast ones
<azonenberg> it will be able to oversample SGMII, 1000base-X, PCIe gen 1, etc
<azonenberg> but only on four channels total
<azonenberg> and we might need to make a differential probe head for it, MEAD is single ended
<bvernoux> but MEAD/CONWAY can do DDR/DDR2/DDR3 ?
<bvernoux> DDR4 is maybe too fast ...
<azonenberg> CONWAY isnt finished, specs are TBD
<azonenberg> the comparators in MEAD are able to run up to 4 Gbps, the limiting factor is how fast the host can sample
<azonenberg> MAXWELL will have ~80 1.25 Gsps channels and 4 10 Gsps channels
<azonenberg> DENNARD will have 20+ 12.5 Gsps channels
<azonenberg> DENNARD is intended to be competitive with the lecroy hda125 which is intended for ddr4 debug
<azonenberg> MAXWELL will be able to decode an entire ddr1 bus including address and data, but probably not anything faster
<azonenberg> it's targeting a different design space, lots of slow signals
* monochroma thinks we need to come up with product model number scheme to identify these platforms based on their features
<azonenberg> e.g. sniffing every pin on a midsized MCU at once
<azonenberg> monochroma: All of these names are internal development codenames
<azonenberg> they're not intended to be used as actual product names
<azonenberg> they're not secret, any more than "sandy bridge" or "nehalem" are
<azonenberg> the rationale being that we need a name for the directory in the repo, the cad files, etc before specs are final
<azonenberg> and the final product name will depend on actual specs
<monochroma> when you have 6+ randomly selected code names for very similar products it's hard to keep track of what's what without refernceing a LUT :P
<bvernoux> the name are quite confusing ;)
<bvernoux> but yes internal name
<_whitenotifier-f> [starshipraider] azonenberg pushed 1 commit to master [+0/-0/±1]
<_whitenotifier-f> [starshipraider] azonenberg fae1da8 - Updated README
<azonenberg> the README in azonenberg/starshipraider, just updated, should have all of the codenames and rough specs
<bvernoux> ha nice README updated
<azonenberg> Anyway, MEAD is at fab now and should be capable of 4 Gbps data rates given a host-side LA that can keep up
<azonenberg> I plan to start design work on MAXWELL soonish
<bvernoux> at the end it is planned to be something autonomous all in one ?
<bvernoux> for the Scope
<bvernoux> or a PC based instrument ?
<bvernoux> like Picoscope ...
<bvernoux> I ask that as for high end stuff I prefer to have something autonomous
<bvernoux> as it is a real mess connecting tons of things on the PC
<azonenberg> bvernoux: all of these instruments are meant to live on a 19" rack and plug into a switched ethernet fabric
<azonenberg> which you can connect to from whatever
<bvernoux> ha ok
<azonenberg> a standalone host could certainly be built but that is not the design goal for the time being
<azonenberg> i'm targeting my use case which is a rack on a bench with dozens of instruments packed into the least rack space possible
<bvernoux> yes it can be added to the rack anyway with a computer unit + screen ...
<azonenberg> exactly
<azonenberg> oh also status wise, einthecorgi2 said they'd work on the active probe stuff. i havent heard updates in a bit, need to check in
<miek> i think PC-based things feel cumbersome right now because there aren't any good ones yet :)
<azonenberg> Yes, exactly
<azonenberg> And we're trying to fix that
<miek> indeed!
<bvernoux> to fix the fact scope take 2minutes too boot now ;)
<bvernoux> what a shame ;)
<azonenberg> yeah i expect my instruments to all be up in the time it takes the fpga to load from qspi flash
<azonenberg> so a few seconds max
<monochroma> ultimately, the most expensive integrated scopes you can buy are... just PC attached scopes as well ;)
<azonenberg> yeah. just an overpriced pc stuck in the same enclosure
<azonenberg> but once you put two on the rack now you have TWO overpriced pcs, two screens...
<azonenberg> which is where you start to waste space
<bvernoux> so far with a light linux distribution with SSD and modern CPU(s) it shall never take more than 15s
<azonenberg> I am actually considering making a rack-mountable UI module we can use with scopehal instruments to provide an all-in-one appliance at some point
<azonenberg> but no immediate plans
<bvernoux> last time i saw a Lecroy demo where it is required to reboot the whole scope just to update the DB it was soo slowwwww
<azonenberg> lolol
<bvernoux> with windows behind ...
<azonenberg> yes
<bvernoux> what a shame
<monochroma> don't forget to update your scopes anti-virus!
<bvernoux> for a 30KUSD Scope ;)
<azonenberg> anyway, so as far as hardware status goes
<bvernoux> even more in fact it was the top scope
<bvernoux> 80GSPS ...
<bvernoux> yes with scope antivirus ;)
<azonenberg> the BLONDEL analog board is gated right now waiting on a) me to finish characterizing and tweaking the AFE prototype and b) einthecorgi2 to figure out the active probe stuff
<bvernoux> do not laugh but we sell server access ... under Linux
<bvernoux> and it is mandatory to add an antivirus ;)
<bvernoux> as certification and so on want an antivirus ;)
<azonenberg> the BLONDEL mainboard is gated because connector placement depends on final dimensions of the analog board, although there's room to do some schematic-level work at this time
<bvernoux> even if it run under Linux box haha
<azonenberg> anyway, so basically no hardware design possible on BLONDEL at this time, but we can do more tests on the frontend
<azonenberg> so the next step on the hardware front, once the v0.9 probe board is done, is to start schematic entry on the MAXWELL mainboard
<azonenberg> which will be the most complex digital board i've ever done
<azonenberg> an entire sodimm of ddr3, the biggest fpga i've ever used in a design, 40G ethernet, 80+ LVDS inputs
<bvernoux> I have just crack opened the DSLogic U3Pro probe ;)
<bvernoux> interesting
<bvernoux> 2 resistor and 1 capacitor ;)
<bvernoux> with a micro PCB circuit
<bvernoux> let's take photo and measure each value ;)
<bvernoux> I also confirm the cable has a shield
juli965 has joined #scopehal
<bvernoux> azonenberg, it could be interesting to build an open source UFL / Balanced Calibration Kit ;)
<bvernoux> the one I bought Megiq is quite expensive for what it is
<bvernoux> some 3D Printed stuff with standard UFL & Pin adapter
<bvernoux> which can work up to 4GHz
<bvernoux> it is interesting mainly to tune PCB and Antenna where there is no space to add a SMA connector (but just soldering an UFL connector or 2 pin connectors)
<azonenberg> the problem with that is that u.fl only has a few mating cycles lifetime
<azonenberg> its rated for like five cycle i think
<bvernoux> Yes but I think more using Ultraminiature PCB pin connectors 1.27mm
<bvernoux> instead of those UFL
<bvernoux> which are defined as 500 cycles
<bvernoux> a bit like SMA
<bvernoux> and it is ultra fast to do calibration
<bvernoux> impact on the PCB is minimum
<bvernoux> depending on frequency
<bvernoux> the must for less impact if UFL ...
<bvernoux> -if+is
<bvernoux> where it is not possible to solder an SMA connector which is a must
<bvernoux> mainly for small RF Antenna embedded ...
Famine- has quit [Read error: Connection reset by peer]
<azonenberg> Also... I ordered some "thread rolling screws", which are apparently slightly different from self tapping screws and meant specifically to go into plastic
<azonenberg> #4-20 thread, 1/4 and 1/2 inch, and some 5mm spacers
<azonenberg> this should be everything i need to assemble the MEAD kit
<azonenberg> MEAD enclosure*
<azonenberg> should be here today actually, gotta check the mail
<azonenberg> i also ordered a set of nearfield EM probes, it was only $200 for four probes and i figured they'd be handy to have for tracking down noise issues etc
<monochroma> which ones?
<azonenberg> Tekbox TBPS01, sold by tequipment but via amazon
<azonenberg> i figured it was cheap enough if they're not awesome its not great loss
<azonenberg> likely better than something i'd homebrew with a loop of wire and a sma
<azonenberg> and i dont feel like buying a $$$$ R&S set or something
<azonenberg> Hmmmm
<azonenberg> So here's a thought
<azonenberg> it shouldn't matter how well impedance matched the probe tip is
<azonenberg> because i'm going to have a massive step right behind the tip where the first 100 ohm resistor is
<azonenberg> I wonder if the current issue is related to the sma connector mismatch again?
<azonenberg> i might want to do some TDR testing or something of the probe to see if that sheds any light on it
<electronic_eel> azonenberg: I have the tekbox nearfield set too, they work well enough for me
<electronic_eel> but nearfield measurement isn't precision measurement, holding angle and similar things influence the results very much
bvernoux1 has joined #scopehal
<electronic_eel> so it works well to guess where an emi issue comes from, but not for something like to decide if you'll pass compliance testing or not
bvernoux1 has quit [Read error: Connection reset by peer]
bvernoux1 has joined #scopehal
bvernoux has quit [Ping timeout: 256 seconds]
<miek> yeah, that set is good. a colleague got it for doing some EMC testing recently - after a half day in the proper test lab, results were marginal so he used that set & an SDR to take relative measurements and make fixes. sailed straight through with plenty of margin the next time
<azonenberg> Sooo this is interesting
<azonenberg> red/pink are S11/S22 of the probe stuck into the open end of a SMA F-F coupler
<azonenberg> blue/cyan are S11/S22 of a thru line on the same stackup with the same SMA
<azonenberg> interesting to note it's not reciprocal
<azonenberg> i'm not quite sure how that's possible as i have no active components...
<azonenberg> this is port impedance
<miek> i guess each reflection is going to be attenuated by the resistors in the path
<azonenberg> no
<azonenberg> this is using 0R's lol
<azonenberg> i'm trying to match the pcb itself
<azonenberg> it should be a thru line
<miek> probably still some loss though?
<azonenberg> note here that port 1 (the SMA coupler side) sees impedance jumping up to ~85 ohms then down to below 20 then up above 140
<azonenberg> port 2 (probe side)... i dont even know what to make of it
<azonenberg> again the thru line on the same stackup shows pretty flat, a tiny bit of oscillation but close to 50 ohms the whole way
<miek> if there's a huge reflection at the probe tip, it'll show up more on S11 vs. S22?
<miek> (assuming non-negligible loss on the line)
<azonenberg> I think so
<azonenberg> I'm going to try doing some time domain reflections on the VNA, that might be helpful i think
<azonenberg> time domain measurements*
bvernoux has joined #scopehal
<miek> yeah, i find those can be super helpful. especially if you can do it in realtime, you can literally poke a finger at a specific point and see it on the plot :)
bvernoux1 has quit [Ping timeout: 256 seconds]
<azonenberg> ok so starting out, this is time domain S11 with the coupler installed and no probe
<azonenberg> and this is with a terminator on the coupler
<azonenberg> this seems to suggest it is the tip
<azonenberg> with a positive reflection aka too high impedance
<azonenberg> and this is with the probe SMA open circuit
<azonenberg> showing the big reflection off that
<azonenberg> So that mismatch is definitely at the tip
<miek> agreed
<azonenberg> and this is one of the pico ta061 probes for comparison's sake
<azonenberg> small hump likely the tip, big the attenuator
<azonenberg> here's tdr of my probe with resistors installed
<azonenberg> so first hump is the tip, second is the attenuator which seems to be electrically short (i dont see individual humps in the response from the resistors)
<azonenberg> this is the thru line now with no probe, let's see what happens if i add a probe mid-span...
<azonenberg> actually let me terminate that to hide the big reflection at the end
bvernoux has quit [Quit: Leaving]
<_whitenotifier-f> [scopehal] noopwafel commented on issue #15: Add Picoscope scope driver(s) -
funkylab[m] has quit [*.net *.split]
funkylab[m] has joined #scopehal
azonenberg has quit [*.net *.split]
futarisIRCcloud has quit [*.net *.split]
sorear has quit [*.net *.split]
wbraun has quit [*.net *.split]
miek has quit [*.net *.split]
sorear has joined #scopehal
wbraun has joined #scopehal
miek has joined #scopehal
futarisIRCcloud has joined #scopehal
Error_404 has quit [*.net *.split]
Pretzel4Ever has quit [*.net *.split]
deltab has quit [*.net *.split]
vup has quit [*.net *.split]
Pretzel4Ever has joined #scopehal
Error_404 has joined #scopehal
deltab has joined #scopehal
vup has joined #scopehal
azonenberg has joined #scopehal
funkylab[m] has quit [Ping timeout: 244 seconds]
futarisIRCcloud has quit [Ping timeout: 246 seconds]
futarisIRCcloud has joined #scopehal
funkylab[m] has joined #scopehal