azonenberg changed the topic of #scopehal to: libscopehal, libscopeprotocols, and glscopeclient development and testing | https://github.com/glscopeclient/scopehal-apps | Logs: https://libera.irclog.whitequark.org/scopehal
Degi_ has joined #scopehal
Degi has quit [Ping timeout: 248 seconds]
Degi_ is now known as Degi
bvernoux has joined #scopehal
<azonenberg> miek: yes
<azonenberg> it's something you can measure and compensate for
<azonenberg> but enig loss can matter for this kind of distance
<azonenberg> especially for something like a vna standard where you care about fractions of a dB
<azonenberg> the ~45mm long 2x thru standard on my directional coupler test board, for example, was oshpark ENIG, no soldermask, microstrip
<azonenberg> it hit 1 dB of insertion loss just past 4 GHz and at 8 GHz was around 1.5
<bvernoux> @miek, I have computed the loss for FR4 it really depends on the trace length and max Freq
<bvernoux> @miek, for FR4 (with Er 4.6) Loss is about 0.3dB/cm @6GHz
<bvernoux> @miek, what is important is to avoid a too small traces width especially < 0.2mm which have very bad effect on the loss
<bvernoux> @miek, for that OSHPark FR408HR is prefect up to more than 8GHz
<bvernoux> @miek, good rule of thumb is provided in https://www.edn.com/loss-in-a-channel-rule-of-thumb-9/
<bvernoux> Attenuation figure of merit: 0.2 dB/inch/GHz, for a lossy channel, 0.1 dB/inch/GHz for a low loss channel.
<miek> how much loss is too much for a cal standard? i know that ideally open/short standards should maximise the reflection for best results, but i don't know how important it is
<azonenberg> miek: no idea. but the loss is not something you can ignore and consider "zero"
<azonenberg> i would imagine that any delta between the actual and modeled loss of the standard is going to directly translate to measurement error
<Degi> How much effort would a cal standard be based on air coax lines? Like soldering two coaxial gold plated copper tubes between two coax connectors
<azonenberg> like these?
<azonenberg> Degi: basically, its doable and is the way to go if you need ultra high grade accuracy
<azonenberg> but 99% of the time is more trouble than it's worth
<Degi> I see, yeah PCBs are much easier to assemble
<bvernoux> miek, for a cal standard we do not care in fact
<bvernoux> miek, as we will de-embedded it after as it will be fully caracterized
<bvernoux> miek, On my side what I want is to use the eCal LibreCAL also as a switch in some case and for that it is important to have something linear from few KHz up to 6GHz (or even later 8GHz)
<bvernoux> miek, Anyway I have computed in worst case the path loss (depending on trace length+width FR4 properties with ENIG) will be less than 1dB
<bvernoux> That does not take into account the switch loss
<bvernoux> So the answer if we will use a better PCB than FR4 with Er 4.6 like OSHPark FR408HR with Er 3.66
<bvernoux> We will have a gain of maybe 0.3 or 0.4dB max
<miek> yeah, of course it'll be fully characterised but surely the quality of the standards still matters to some extent
<bvernoux> So it will be better but clearly who cares about 0.4dB especially as OSHPark PCB you will need to assemble all yourself
<bvernoux> I have chosen the best possible RF Load already used by @azonenberg
<bvernoux> So I do not have lot of doubt especially to reach up ton 8GHz easily without any strange things
<bvernoux> At the end the performance will mainly depends on the switch (if the Impedance of the PCB traces are respected +/-10% with Z=50 Ohms)
<bvernoux> On my case I will do all at JLCPCB
<bvernoux> as they will do also the assembly of the 5 units with the components I have bought
<bvernoux> Anyway I will do all measurements when I receive the boards to check everything and also see the variations over 5 boards
<bvernoux> It is what is missing on the LibreCAL GitHub as there is no any S2P files
<bvernoux> or S1P files
<bvernoux> In my case I will do full 2 ports S2P files with S11/S21/S12/S22
<bvernoux> with my VNA HP8753D which goes from 30KHz to 6GHz with 1601pts
<bvernoux> Of course the HP VNA will be fully/freshly calibrated with my expensive Calibration Standard before
<miek> nice. yeah, i've been following the progress on github but it'll be great to see some measurements
Bird|otherbox has quit [Ping timeout: 260 seconds]
<bvernoux> @miek, I have done some measurements on lot of experimentations with stripline/TRL
<bvernoux> Up to 6GHz which is the limit of my VNA so far
<bvernoux> With some other stuff I have not pushed but I will finish with the eCal(LibreCAL) as doing so much test will kill my Calibration Standard/VNA Cables(which are very expensive) I have already planned to buy new VNA SMA cables to replace my older ones
<azonenberg> bvernoux: i've been eyeing some Koaxis cables for VNA applications, if you are interested in trying them out
<azonenberg> i've used a lot of their other cables and they're excellent
<bvernoux> azonenberg, ha yes maybe as so far I just need to buy a 2nd https://www.mouser.fr/ProductDetail/139-FLC-2FT-SMSM%2b
<bvernoux> Which are good cables
<bvernoux> But maybe the ones from Koaxis are even better
<azonenberg> although i dont see KF210 in their online cable configurator anymore. maybe it's out of stock right now?
<azonenberg> slightly lossier than KF210 but skinnier and more flexible
<azonenberg> My high grade cables I use for full BW measurements on the 16 GHz scope are KF120
<azonenberg> insertion loss of only 1.1 dB for a 2-foot cable at 16 GHz
<bvernoux> Yes they seems very good
<bvernoux> and price similar to MiniCircuits FLC-2FT-SMSM+
<bvernoux> Which is what I select for my VNA so far
<bvernoux> In my case I do not need very long cable just something max 60cm
<azonenberg> I currently use CBL-2FT-SMNM+
Bird|otherbox has joined #scopehal
<azonenberg> which are nice cables but i think koaxis are even nicer
<bvernoux> Yes to be checked
<bvernoux> I'm very tempted to buy 2xKoaxis
<bvernoux> 60cm each
<bvernoux> with KF120
<bvernoux> Phase matched
<bvernoux> If they are phase matched I suspect they are also very nicely Delay matched
<bvernoux> as they provide 2 choices
<azonenberg> yeah
<azonenberg> its just what the pass/fail characteristics are
<bvernoux> Koaxis KF120 have quite identical Insertion Loss vs MiniCircuits CBL-2FT-SMNM+/FLC-2FT-SMSM+ at same length
<azonenberg> Yes KF210 is lower loss but seems to be out of stock right now
<azonenberg> since it's listed on the site but you cant order it
<bvernoux> On my case I do not search the lowest insertion loss but also robustness
<bvernoux> and they do not provide any details on max number of connection
<azonenberg> Koaxis cables are generally very well built and come with (on request) serialized s-parameters
<azonenberg> i would expect mating cycle endurance to be on par with competitive cables
<bvernoux> Do you have checked Junkosha MWX7 series cables ?
<bvernoux> They are clearly amazing
<bvernoux> But also much more expensive
<bvernoux> about 250USD minimum per cable
<bvernoux> They have amazing phase stability I think they beat all other cables
<bvernoux> They have amazing spec with Flex Test too
<azonenberg> yeah i havent used junkosha but i've heard good things
<bvernoux> The FLC series are tested => Performance qualified to 20,000 fl exures
<bvernoux> I have it in my Mouser Shopping basket as I have already one that I have never used as I was waiting to buy the same later
<bvernoux> it is EMI Gaskets, Sheets, Absorbers & Shielding 80 - 115dB D Shape L-22" Shield Gasket
<bvernoux> For the LibreCAL
<bvernoux> I was planning to buy some additional connector saver like that SF-SM50+ but it is out of stock
<azonenberg> woo good news from lecroy
<azonenberg> the 4 GHz differential probe i sent in with a suspected bad cable actually had a failed connector
<azonenberg> which was covered under warranty so they fixed it at no cost
<bvernoux> ha great !!
<bvernoux> What was the result with that probe it was totally usable ?
<bvernoux> unusable
<azonenberg> So the way LeCroy's WaveLink system works is they have the "platform" which has a bunch of power supplies, dacs, etc
<azonenberg> to supply power and bias / offset voltages
<azonenberg> and then the amplifier mates with it
<azonenberg> One of those power connections was intermittently failing open when you bent the cable a certain way
<bvernoux> ha yes very bad
<azonenberg> and the amplifier output would peg at -Vdd because it wasn't getting a positive supply
<bvernoux> yes can be crazy to debug that
<azonenberg> it would work as long as you held the cable at exactly the right angle :p
<azonenberg> but yes was useless for practical work
<azonenberg> the amplifier was quite old and i was ready to just declare it a write-off as it was an older model and i didnt want to pay for a repair
<azonenberg> so i sent it in on a "let me know before you fix it" basis
<azonenberg> turns out while the amplifier was old, the platform was fairly new (from 2020)
<azonenberg> and under warranty until october of this year
<azonenberg> So free repair
<azonenberg> My 16 GHz scope is done and shipping back shortly as well
<azonenberg> at least, as soon as I pay for the upgrade
<azonenberg> Which is a bit tricky because i had to cancel the credit card they have on file due to some fraudulent charges recently :p
<azonenberg> So i'm in the process of sorting all of that out
<bvernoux> What do you have upgraded on the 16GHz scope ?
<bvernoux> Upgraded it to >20GHz ;)
<azonenberg> Swapping the processor board out from the old core2quad running vista to a modern i7 running win10
<azonenberg> swapping the analog system out from the original 8Zi platform to the newer 8Zi-A which comes with a newer frontend, i think like 30% lower noise
<azonenberg> replacing the 14 year old touchscreen which would often register touches at the wrong spots
<bvernoux> ha great
<azonenberg> and then a few other random odds and ends like a missing screw on the front panel
<bvernoux> the i17 could maybe run glscopeclient ;)
<azonenberg> Yes, it definitely could
<azonenberg> anyway so at some point in the indefinite future I will be sending it back to upgrade to an 8Zi-B, the current platform, but that was too expensive to do right now
<bvernoux> Main issue I think is there is no GFX card so using HD7000 or similar of the CPU could be a bit limited
<azonenberg> The big benefit from the base to the -B is that the B can interleave channels to get you 2x 80 Gsps
<bvernoux> Ha yes clearly amazing feature
<azonenberg> and then i can also bump max memory from 128 to 256M points (or 512M if interleaving)
<azonenberg> or bandwidth from 16 to 20, 25, or 30 GHz
<azonenberg> but i wanted to get it on the current hardware platform before doing any of those upgades
<bvernoux> I imagine does they have asked if you want to buy software options ;)
<azonenberg> No they didnt
<bvernoux> And you told them no I will use glscopeclient ;)
<azonenberg> if i was buying a new scope maybe, but this was just done through the service department
<azonenberg> but it's no secret that i use scopehal
<azonenberg> i've met the service engineer at lecroy who worked on my scope last time i sent it in for repair
<azonenberg> he knows about scopehal and has actually mentioned it to other customers who were trying to do multi scope sync stuff
<bvernoux> ha great
<bvernoux> page22
<azonenberg> oh that silly periodic table?
<bvernoux> The items are disposed exactly like periodic table yes ;)
<bvernoux> Too fun
<azonenberg> marketing nonsense. but they are nice scopes lol
<bvernoux> Do you have some option to do eye pattern for SerDes/DDR... on yours ?
<bvernoux> Just to compare with glscopeclient
<bvernoux> On an other subject I'm working with some well know security companies to do an HydraDancer ;)
<bvernoux> All will be opensource and based on WCH CH569
<bvernoux> It will clearly beat anything which exist in term of price/features
<azonenberg> yes the SDA comes with eye pattern and such features standard
<bvernoux> So if anyone are interested to help as there is tons of things to do on embedded side and host side(libusb...)
<azonenberg> the SDA is basically a bundle of the base WaveMaster scope plus th serdes analysis software package, the CDR trigger board, and increased memory depth
<bvernoux> ha great as lot of others like Keysight/Tek provide such features not freely and not as standard
<azonenberg> i mean, if you buy a wavemaster then you have to buy those options
<bvernoux> ha ok
<azonenberg> the sda is basically just a package deal for getting them all at once
<azonenberg> and for some reason it's a new SKU
<bvernoux> Yes yours was intended to be refurbished with 2 years warranty ?
<azonenberg> on the 16G scope? there wasnt much of a warranty, i bought it third party
<azonenberg> i think 30 days
<azonenberg> i was willing to accept the risk of something expensive going wrong because it was such a huge discount vs a new one
<azonenberg> even after all of these upgrades and fixes it's going to cost me like half of a new one
<bvernoux> Yes it is clearly a big deal even if there is some risks
bvernoux has quit [Quit: Leaving]