<azonenberg>
they're in the queue to look at once I get the AKL-PT2 stuff moving
<d1b2>
<david.lenfesty> sweet
_whitenotifier-5 has joined #scopehal
<_whitenotifier-5>
[scopehal] azonenberg pushed 1 commit to master [+0/-0/±2] https://git.io/JYaA7
<_whitenotifier-5>
[scopehal] azonenberg 031101d - LeCroyOscilloscope: correctly detect SDA/WaveMaster 8Zi series as 4 channels, not (last digit of bandwidth in GHz)
<azonenberg>
Looks like 23.88 mm port extension for the new test fixture
<monochroma>
azonenberg: have you playd with the detachable front panel at all? :P
<azonenberg>
I detached it to look, but havent actually used it in the detached state
<azonenberg>
my bench is full enough of stuff
<azonenberg>
i want stuff on the rack out of my way :p
juli969 has joined #scopehal
<_whitenotifier-5>
[starshipraider] azonenberg pushed 1 commit to master [+1/-0/±1] https://git.io/JYVi3
<_whitenotifier-5>
[starshipraider] azonenberg 52cfbce - Initial release of AKL-PT2 datasheet/manual
<bvernoux>
azonenberg, does glscopeclient already support it ?
<bvernoux>
azonenberg, I suspect you need to do some modifications in glscopeclient to support the SDA816Zi
<azonenberg>
as of 031101d in libscopehal, it functions
<azonenberg>
But i dont have all of the memory depth/sample rate tables done, or support for the serdes trigger
<bvernoux>
ha yes nice few lines of code to basically support it
<bvernoux>
it seems the 4 chan are in // SMA & BNC
<bvernoux>
what happen if you use BNC & SMA at same time does it act like a bad mixer ;) ?
<azonenberg>
No there's a relay to mux
<bvernoux>
ha ok
<bvernoux>
I was checking KeySight scope ;)
<bvernoux>
they are totally crazy in price for similar perf
<bvernoux>
without any option ...
<bvernoux>
.not even competitive vs Tek ...
<bvernoux>
Their commercial told me to check R&S Oscilloscope ...
<bvernoux>
If I want an oscilloscope which can be upgraded from 1GHz to 8GHz (in SW) starting at 20KUSD ;)
<bvernoux>
as Keysight as clearly nothing in that range...
<bvernoux>
their proposal was a 16GS/s 4 Chan ...
<bvernoux>
for < 100KUSD
<bvernoux>
or else their proposal is their university oscilloscope(toys) with less than 1GHz BW 5GSPS ...
bvernoux has quit [Read error: Connection reset by peer]
<azonenberg>
Ok so, I did a bunch of measurements of AKL-PT2s and all of my other probes with LeoBodnar's pulse generator (BNC, the SMA one seems to be acting flaky for some reason and i havent troubleshooted yet) and the 16 GHz scope
<azonenberg>
https://www.antikernel.net/temp/old-v0p1-probe.png this is the original solder-in probe (20x, much shorter than the PT2). Stupidly fast rise time but terrible flatness with ripples and spikes all over the place
<azonenberg>
the time domain waveform shows some huge dips and overshoot (this is also a different time scale than the rest for some reason)
<LeoBodnar>
what's the deal with SMA pulser problems?
<azonenberg>
LeoBodnar: Dont know yet. I torqued the connector to spec and got spikes on the rising and falling edges, looked like it was AC coupling or something
<azonenberg>
as i loosened it, i got a short burst of good looking signal
<azonenberg>
I haven't had a chance to investigate in any detail
<LeoBodnar>
let me know if you find anything
<azonenberg>
(This is the one that i swapped the SMA on after it failed the pin position measurement)
<LeoBodnar>
yeah I remember
<azonenberg>
Anyway, continuing down the list of probes...
<azonenberg>
slow edge, but nice and smooth with no artifacts whatsoever
<azonenberg>
https://www.antikernel.net/temp/pico-ta061.png Pico TA061 1.5 GHz transmission line probe - smooth-ish, but horrid overshoot (it continues ringing beyond the time scale of this capture)
<azonenberg>
https://www.antikernel.net/temp/pico-921.png PicoConnect 921 transmission line probe. Definitely slower rise than my probe, although it's a smoother rising edge that then rings a bit vs my little step
<azonenberg>
https://www.antikernel.net/temp/pt1-v1p3.png AKL-PT1 v1.3 prototype. This is the one TSP reviewed, you can see the big hump on the rising edge then the sharp dip after. That's the time domain view of the 5 GHz peak i think
<azonenberg>
https://www.antikernel.net/temp/pt1-v1p4.png AKL-PT1 v1.4 prototype. This has more overshoot because I removed the filter to study the tip overshoot without interference
<azonenberg>
https://www.antikernel.net/temp/d420-si.png LeCroy D420 active probe with the solder-in (SI) tip. This is a 4 GHz active diff probe, you can see it's quite a bit lower BW than the PT1 or PT2. It also has noticeable overshoot
<azonenberg>
https://www.antikernel.net/temp/d1330-pt.png LeCroy D1330 active probe with the handheld (PT) tip. This is a 13 GHz active diff amplifier, but limited to 10 GHz with this tip
<azonenberg>
I've also got a bunch of probes that slightly failed spec during my production testing, i promised a few to friends already but i can let more of my "dings and dents" probes go to folks here if i continue to have not-great yield
<azonenberg>
So far of the... 22 I think? that i've assembled, one was destroyed by a stupid slip of the hand during characterization ripping the tip off, two are on a DUT but not fully characterized, which leaves 19 characterized live probes
<azonenberg>
12 of those are in sellable condition so i think that means i have 7 that failed spec somewhere that are still perfectly usable ~5 GHz probes
<azonenberg>
but with worse flatness or bandwidth than the datasheet limits
<Degi>
I guess if the yield continues to be 50 % you could sell a lower specced version in terms of flatness etc.?
<azonenberg>
These first few are being given away free to friends and coworkers, but if i have a bunch more bad ones on the next production batch i think i'll probably offer them to anyone in the channel willing to pay BOM cost and shipping
<azonenberg>
Degi: i could, but i'm hoping i can tighten up the tolerances a bit :p
<Degi>
Heh yes, that would be good
<Degi>
Also how is the durability? Does the PCB delaminate after being reused too often?
<azonenberg>
if you overheat it maybe? so far i have not had any fail from soldering
<noopwafel>
how much *is* BOM cost approximately for these?
<azonenberg>
the only one i killed was from the tape coming loose as i moved the VNA cable around
<noopwafel>
(I remain on the side of 'you should sell these as lower-specced' also)
<azonenberg>
and tearing the probe off the board leaving the castellations behind :p
<azonenberg>
noopwafel: BOM is roughly $25. Boards were about $1250 after shipping and import duty for 250 pcs so $5
<azonenberg>
then a $12ish SMA and three $2 resistors
<azonenberg>
then a bit of epoxy and solder paste
<azonenberg>
but assembly/test is pretty labor intensive, especially since they're 100% VNA tested
<Degi>
Are the PCB materials similar to the ones used by OSHPark in terms of RF performance?
<azonenberg>
This is the same flex substrate used by oshpark
<azonenberg>
but coverlay instead of soldermask, and impedance control and stiffeners added
<azonenberg>
you can make the same gerbers on oshpark flex and it will work with slightly reduced performance, but you'll have to glue something to the board under the SMA in order to solder the connector properly
<azonenberg>
and you'll probably want to stiffen under the resistors somehow to keep the solder joints there from cracking
<azonenberg>
I did all of my prototyping at oshpark though
<Degi>
Hm yes, I was wondering because FR4 is rather bad for RF and I heard somewhere that flex is a cheap replacement without using RO4350B
<azonenberg>
The AKL-PT1 prototypes all use oshpark FR408HR
<azonenberg>
which is way better than standard FR4
<Degi>
Hm good to know
<azonenberg>
And the reason I like 4350B specifically, vs any of the other rogers or other high speed laminates, is that it's very close in performance otherwise to FR408HR
<azonenberg>
lower loss but almost the same Er, available in the same thicknesses generally, etc
<azonenberg>
So i can prototype on oshpark and run the same gerbers on rogers and get the same result except for loss
<Degi>
Hm, I think I could accept like 6 dB on 10 cm at 6 GHz#
<azonenberg>
yeah it totally depends on what you're doing
<azonenberg>
i'm making probes so i'm pushing hard for every 0.1 dB
<azonenberg>
and more importantly, i want to get the flattest response possible
<azonenberg>
this is a much more challenging problem than making something with low loss in, say, the 5 GHz wifi band
<azonenberg>
the broadband flatness is tricky
<Degi>
Hm yes, I think in my application I could just calibrate the flatness / loss up to a few dB
<azonenberg>
Yeah. like that auburn instruments probe made out of a ballpoint pen
<azonenberg>
auburn technology*
<Degi>
Huh? Interesting
<azonenberg>
i talked to an engineer at signalhound who uses it
<azonenberg>
the advertised flatness graph is... optimistic to say the least
<azonenberg>
he said there was like an 8 dB range with multiple resonances throughout the band
<azonenberg>
but that it worked well for narrowish band applications if you used a signal generator to calibrate the gain at a given frequency
<Degi>
8 dB flatness? oof
<azonenberg>
Yeah :p
<Degi>
Isnt that like 2.5 : 1 voltage ratio?
<azonenberg>
it's substantial yes
<azonenberg>
i guess the theory is that for narrowband work you dont really care about the gain at other frequencies
<azonenberg>
you measure the gain at your frequency of interest and calibrate it out
<azonenberg>
and if a GHz away to either side it has 3 dB more or less gain, you don't care
<azonenberg>
but for signal integrity or other broadband applications, flatness matters a lot
<Degi>
Huh, the signalhound analyzers look neat
<azonenberg>
the specan? yeah, the SM200C is on my lab wishlist
<azonenberg>
I have 10GbE everywhere already, it seems like a perfect addition to my fleet
<azonenberg>
Unrelated, something is weird about my D400A-AT-PB2 probe and this scope. I'm not yet sure what is the cause
<azonenberg>
but it seems to always read flatline on the SDA
<azonenberg>
the same probe works fine on the waverunner, and my other ProBus probes work fine on the SDA
<azonenberg>
I suspect it might be the WL-PBUS2 being too new for the firmware on this older model scope
<azonenberg>
or it might have some hardware incompatibility due to age
<azonenberg>
(if so, upgrading to the 8Zi-B series should fix that)
<monochroma>
azonenberg: oh yeah, i think i saw the answer i skimmed in the backlog, but wanted to ask earlier, so it has BMA and BNC input jacks, and they just RF relay switch between a channel using BMA or BNC?
<azonenberg>
Yeah
m4ssi has joined #scopehal
bvernoux has joined #scopehal
bvernoux has quit [Quit: Leaving]
<d1b2>
<mubes> @azonenberg is there a code prettifier anywhere or spec for what you want to see in the repository?
<d1b2>
<mubes> Grr...I thought I'd looked for a clang-format. Ta!
bvernoux has joined #scopehal
<azonenberg>
bvernoux: did you see my rise time comparison tests from earlier?
m4ssi has quit [Remote host closed the connection]
<Famine_>
ok i really want to throw a brick at who ever is writing hiroses documentation... there is no max board thickness in the catalog for mid mount usb c connectors, its not in the drawing, so you have to click through 4 drop menus on the hirose site and check each connector
<d1b2>
<mubes> Performance is pants, 1wf/s at 1Msamples, but it's reasonably stable...just got to be careful with making sure the buffers a clear before you do anything else. I'll tidy it up as 'sig2' which is what it's called at the moment and then you can decide on the best way to bolt it in...the obvious way is probably for the original siglent driver to instantiate this or the LeCroy driver depending on which scope it finds itself connected to. There is
<d1b2>
some support for the digital channels but it's totally untested. I guess this is a shout out for some singlet alpha testers to give it a whiz.
<azonenberg>
The current siglent driver is basically garbage afaik
<azonenberg>
it doesnt really work
<azonenberg>
honestly i'd replace it outright
<azonenberg>
it's more likely to work than what's there now :p
<d1b2>
<mubes> Well, that certainly makes things easy :-)
<d1b2>
<mubes> That's for tomorrow now anyway. I'm off to bed before I break it again.