azonenberg changed the topic of #scopehal to: libscopehal, libscopeprotocols, and glscopeclient development and testing | https://github.com/azonenberg/scopehal-apps | Logs: https://freenode.irclog.whitequark.org/scopehal
Degi_ has joined #scopehal
Degi_ is now known as Degi
_whitelogger has joined #scopehal
juli968 has joined #scopehal
sam210723 has joined #scopehal
juli968 has quit [Quit: Nettalk6 - www.ntalk.de]
juli968 has joined #scopehal
<azonenberg> miek: veeeery interesting data coming off these new probes
<azonenberg> I'm going to have to re-measure some of the old ones i think
<azonenberg> overall shape is about the same but the ripples in 2-3 GHz are almost gone
<azonenberg> And there's a big dip around 3.3 GHz in about half of them
<azonenberg> the rest have a peak in about the same spot
<monochroma> :o
<azonenberg> I need to get to the bottom of this before going any further
<azonenberg> the new data looks "better" but i dont know why
<azonenberg> and if i re-measure and it looks better i need to know why
<azonenberg> one thing i did change for testing A69 was fully torque the SMA onto the fixture
<azonenberg> it's possible hand tightening it is responsible for the ripples
<azonenberg> i'm gonna try redoing A68-004 with hand tightening then fully torqued and see what that changes
<azonenberg> the other question is why the dips in the A69 probes are so different
<azonenberg> not sure if soldering variation or probe to probe variation or what
<azonenberg> also going to write up a more formal test procedure to try and eliminate some of this variance
bvernoux has joined #scopehal
<azonenberg> Aaaand seems conclusive
<azonenberg> those ripples were mismatch at the SMA from not torquing it enough
<monochroma> oh cool
<azonenberg> So now i have to re-cal all of the A68 probes now
<azonenberg> as those measurements are invalid
<azonenberg> lol
<bvernoux> haha interesting
<azonenberg> same probe, same fixture
<azonenberg> hand tight and properly torqued in blue and red respectively
<bvernoux> you was not doing SMA torquing with SMA Torque Wrench with 0.45 Newton-meters (3.98 inch-pounds) ?
<azonenberg> I didn't do it on the A68 probes (prototypes), I did on A69 (production lot)
<azonenberg> And this appears to explain the variance I'm seeing
<bvernoux> yes clearly
<azonenberg> So now i have to re-test A68
<bvernoux> on VNA is see big differences also when I'm not using my SMA Torque wrench
<azonenberg> only four probes, not a big deal
<bvernoux> yes
<bvernoux> in your picture the red curve was not done with SMA torque wrench ?
<bvernoux> as there is big issue near 3.3GHz ;)
<azonenberg> No
<azonenberg> Red was properly torqued, blue was hand tight
<azonenberg> I think that dip is real
<bvernoux> ha ok it is S11 in fact
<azonenberg> although the height of the dip seems to vary from probe to probe. Not yet sure on cause
<azonenberg> thoughts?
<bvernoux> It is first time is see connector gaging check ;)
<bvernoux> nice
<azonenberg> I'm trying to make my results as transparent and reproducible as possible
<bvernoux> yes nice
<bvernoux> you are soldering each probe to test them ?
<azonenberg> Yes
<azonenberg> Standard edition probes will get a quick risetime check only
<azonenberg> Pro edition will get full VNA characterization across open and 50 ohm load
<azonenberg> actually i might still VNA the standards at least to check S11, we'll see
<bvernoux> what is not clear is step 5
<bvernoux> when you connect terminator to SMA
<azonenberg> there's only one SMA on the probe, i didnt think a picture was needed
<bvernoux> do you mean it is 50ohm terminator on both SMA on test fixture ?
<azonenberg> No. Terminate the probe, VNA to SMA on fixture
<bvernoux> on picture we see the both SMA on test fixture are open
<azonenberg> that's not a thru line, it's an open
<bvernoux> ok
<azonenberg> see the "6mm open" text next to it
<azonenberg> probe is soldered across the end of the open
<bvernoux> ha yes ;)
<bvernoux> I was confused by "Connect terminator to SMA on probe"
<bvernoux> it just means you connect the SMA from probe to the VNA
<bvernoux> Port1
<bvernoux> I think you could avoid to solder the probe on test fixture
<bvernoux> and just measure the probe with VNA
<bvernoux> and so you take a reference and compare with that one
<bvernoux> that will avoid to solder anything
<bvernoux> especially for a test with open
<azonenberg> I'm measuring input impedance of the probe
<azonenberg> I need the castellation connected to something
<azonenberg> i don't trust any kind of mechanical contact to be firm enough for s-parameter measurements
<azonenberg> The SMA on probe is *not* connected to the vna
<azonenberg> the single vna port used for this test goes to the fixture
<azonenberg> which is soldered to the tip of the probe
<azonenberg> the SMA end of th eprobe is terminated
<azonenberg> I'll add this pic to the procedure
<bvernoux> ha ok
<bvernoux> it is more clear like that ;)
<bvernoux> now I understand why you need to solder it
<azonenberg> Yeah. Even more important for S21 measurements
<azonenberg> which are going to be done the same way except across a termination instead of an open
<azonenberg> bvernoux: also not sure if you were here when i said it or not, but my 16 GHz scope is done being repaired and running on cal overnight
<azonenberg> probably won't be done by the weekend so they expect it to ship next week
<azonenberg> I'm quite excited to see just how fast the edges i can get through this probe are
bvernoux1 has joined #scopehal
bvernoux has quit [Ping timeout: 245 seconds]
bvernoux1 has quit [Client Quit]
bvernoux has joined #scopehal
<azonenberg> ... nooooo
<monochroma> D: !
<azonenberg> Nine probes worth of s-parameters are worthless
<monochroma> :O
<azonenberg> That big dip
<azonenberg> it's not real
<azonenberg> My fixture is resonating or something
<azonenberg> this is s11 of the open circuit on the fixture with no probe connected
<monochroma> how did you find it's not the probe?
<azonenberg> i took the probe off the fixture and vna'd the fixture :p
<azonenberg> and it didnt go away
<monochroma> ohhhhhhh
<monochroma> i thought you had already characterized the fixture?
<azonenberg> i was trying a different one that was easier to solder to for this purpose
<azonenberg> Mistake
<azonenberg> this fixture is also less lossy
<azonenberg> so it should have delivered better results
<azonenberg> ok now i am even more confused
<azonenberg> I'm seeing a giant spike with the vna cable floating and nothing hooked up
* azonenberg recalibrates vna
<azonenberg> Mystery solved
<monochroma> vna cal issue? :<
<azonenberg> Yes and no
<azonenberg> The connector saver between the SMA-N cable and the VNA front panel had loosened over time
<azonenberg> i torqued the cable onto the saver but didn't re-torque the saver because its a different size that my torque wrench won't fit
<monochroma> ahhhh
<azonenberg> The saver is, i guess, about a quarter wavelength long at ~3 GHz...
<monochroma> heh
<azonenberg> So my data is indeed invalid, but the fixture was not to blame
<monochroma> does the saver have a flat for a torque wrench? in the WISP industy i'm used to the N connectors just being knurled without any flats for hand tightening only >_>
<azonenberg> These have nuts for tightening them. but it's the wrong size for my wrench
<azonenberg> i have a normal wrench that fits them which i used to fix it
<azonenberg> but it's not properly torqued which i'm not thrilled about
<azonenberg> i should measure the nut and get a proper torque wrench just for the saver
<azonenberg> you know, this calls my older S21 measurements into question as well
<azonenberg> i recall a bump around that freq...
<azonenberg> ooook so now that THAT problem is found, gonna take a break before repeating the whole night's measurements :p
<azonenberg> i found something that looks good from fairview microwave but i think i will have to special order it
<azonenberg> they have different combinations of preset torque and head size
<azonenberg> the heads seem to be changeable
<azonenberg> but they dont ship a 18mm one with 8.85 lbf-in
<miek> would that have shown up by doing a sweep on the open fixture with no probe soldered on? if so, that'd be an easy check to add to the test plan
<azonenberg> Yes
<azonenberg> I'll add that
<azonenberg> anyway, so all the data i sent you before is now suspect
<azonenberg> I found a few other small procedural issues that i've since corrected too
<azonenberg> trying to be as consistent as possible
<azonenberg> far more nice looking data
<azonenberg> note that the spikes there are like 0.1 to 0.2 dB high now
<miek> nice
<azonenberg> I'm only doing the open circuit measurement for now
<azonenberg> the 2-port measurements across the 50 ohm load are on hold until i get the new scope
<azonenberg> since i want to do rise time on the same solder joint
<azonenberg> and not have to put too many thermal cycles on the probe prior to shipping
<azonenberg> miek: (can you tell i'm excited about the new scope? lol)
<azonenberg> i've only been waiting like a moooonth...
<miek> :D
<azonenberg> its extra annoying because i've been down to one scope
<azonenberg> because i traded in the 2 GHz scope already
<azonenberg> expecting the new one to ship right away
<azonenberg> the dealer is upset too, he thought the scope was working :p
<azonenberg> instead he's gonna be paying $9K in repair costs
<azonenberg> not sure what he bought it for but it wouldnt surprise me if he ends up taking a net loss on it
<monochroma> yaknow, i sold an identical model scope on ebay that i found in a muddy ditch on the side of the road a few months back...
<azonenberg> lol
<azonenberg> Much better
<monochroma> :D
<monochroma> what's the spike at ~2.8ghz?
<azonenberg> Dont know. There's one at 1.9 and 4.2ish as well
<azonenberg> could be fixture effects i didnt fully de-embed, could be something in the probe design
<azonenberg> they're <0.2 dB p-p so i'm just gonna ignore them :p
<azonenberg> did a bunch more
<azonenberg> That's what i like to see
<azonenberg> they're diverging a bunch past 3 GHz but still like +/- 0.1 dB consistency
<miek> very nice
<azonenberg> This isn't even half of the total batch i have to do
<azonenberg> but i stopped to crunch the numbers and see how wide the pread was
<azonenberg> it's looking quite nice
<azonenberg> ok, got a bunch more done. Looking at S11 phase I see a +/- 8 degree spread at 6 GHz
<azonenberg> The probe is 925 ps delay and 160.3 mm in length
<azonenberg> So our velocity factor is about 0.578 (which sounds slow, but we'll use it for now)
<azonenberg> Wavelength of 6 GHz in free space is 49.97 mm, scaling that by the velocity factor gives a wavelength on the probe of 28.88 mm
<azonenberg> 8 degrees is 0.022 of a cycle
<azonenberg> Which means our difference in S11 path length is +/- 640 μm
<azonenberg> Sounds plausible for a combination of resistor misalignment (due to soldering drift) and slight variations in placement of the probe on the fixture
<azonenberg> Pretty cool that the VNA measurement has the resolution to see sub-millimeter variation like this
<azonenberg> intuitively, i wouldn't have expected that with only 6 GHz BW
maartenBE_ has quit [Ping timeout: 246 seconds]
maartenBE has joined #scopehal
Error_405 is now known as Error_404
juli968 has quit [Quit: Nettalk6 - www.ntalk.de]
<azonenberg> marshallh: bad stitch or was the board designer high?
<marshallh> someone who makes cursed photohops
<azonenberg> the more i look the worse it gets
<azonenberg> lol
<marshallh> lol yes
<marshallh> used my z0 probe talking to a realtek FAE the other day
<marshallh> nothing like slapping the 20mhz BW limiter on 1-3ghz probes
<azonenberg> lol
<azonenberg> Did you see the AKL-PT2 datasheet draft yet btw?