<azonenberg>
good thing is, i built standards into my characterization board. which are almost certainly a lot better than the pieces of junk that came with the xavna
<azonenberg>
but i still dont exactly trust this thing
<azonenberg>
With good reason lol
<azonenberg>
i honestly think i'm better off not cal'ing it right now
<azonenberg>
so basically at the end of this test i have even less confidence in the xavna than i did two hours ago
<azonenberg>
and i've learned ~nothing about my probe
<azonenberg>
Lo
<azonenberg>
Lol
<miek>
ouch
<azonenberg>
i mean i guess for $299 i can't expect THAT much lol
<miek>
yeah, i used to use a minivna which i think cost similar (probably even cheaper construction) - it was good enough to sweep an antenna and see where it was resonant, but i didn't really trust absolute numbers from it
* azonenberg
continues to drool over the picovna
<azonenberg>
which seems like by far best bang for buck
<azonenberg>
if you want more than "toy" performance
<miek>
i'm a fan of getting lucky on ebay, but it's not the most robust plan :D
<azonenberg>
well the big thing is, i physically cannot fit a big old school vna on my bench
<azonenberg>
i dont have the sapce
<azonenberg>
space*
<azonenberg>
i already have a reflow oven, two large DSOs, two power supplies, two DMMs, two 24" monitors, a keyboard and mouse, a microscope, and soldering station
<azonenberg>
how much more can you fit?
<azonenberg>
getting lucky on an ebay'd picovna i would consider
<azonenberg>
but some massive 50 pound monster? not happening
<azonenberg>
even if you gave it to me free i'd have nowhere to put it
futarisIRCcloud has joined #scopehal
<miek>
yeah, fair enough - it's kinda silly that my one's still so big. most of the inside is half-empty PC case
<azonenberg>
i see a picovna on ebay for $5995 new
<azonenberg>
but not sure why i'd do that
<azonenberg>
not when i can get it for $4911 from tequipment new :p
<miek>
heh
<azonenberg>
So unless somebody shows me something better, as soon as i can afford the picovna my plan is to buy it
<azonenberg>
i mean technically i can afford it now, but if i buy it then i won't be able to pay the construction guys who are coming to fix my flooding problems in may or june
<azonenberg>
and i figure a lab with no vna, but not underwater, is better than a lab full of water that has a vna floating somewhere in it
<miek>
yeah, i don't see any IP rating on it :p
<azonenberg>
my scopes are IPv6 rated, does that count? :p
<azonenberg>
anyway, i'm going to say at the moment that R&D on the current probe pcb is halted
<azonenberg>
because it's better than my instruments can measure
<azonenberg>
i can't improve it because i can't see the flaws
<azonenberg>
sound correct based on the data i've linked?
<azonenberg>
i'll be still doing the respin in a week or two with the repositioned ground lead and tip socket, but that should not have any significant change on rf performance as the coplanar waveguide itself isn't changing
<miek>
yeah, i'd agree with that
<azonenberg>
the probe characterization fixture is at fab now and should ship any day now, the AFE prototype is at fab
<azonenberg>
So next step is the hmcad1520 board?
<miek>
yup :)
<azonenberg>
3 in stock at digikey, lol
<azonenberg>
Hmmmm
<azonenberg>
For the characterization board, thoughts on just using the fpga's internal pll?
<azonenberg>
actually nvm i dont have enough ios on the connector
<azonenberg>
ok so i'm going to provide a clock mux and your choice of a 625 MHz or 1 GHz clock depending on 8 or 12 bit mode
<azonenberg>
i don't want to waste time doing a PLL config on this board
<azonenberg>
i'll build a separate test platform if i have any doubts as to how it will work
<azonenberg>
ok so a few more tidbits from testing: 335 ps risetime and no overshoot for the same test setup through a lecroy zs1500 active probe
<azonenberg>
i'm going to see what happens if i put a compensation cap on this probe
<azonenberg>
See if i can get a smooth, even, fast rise time without a lot of overshoot
<azonenberg>
Let's start with 200 fF
<azonenberg>
200 fF cap slows the rise time down to 185 ps but doesn't eliminate the overshoot
<azonenberg>
It's reduced though
<azonenberg>
let me try 400. Might be worth having a slightly slower probe with less overshoot, at least for now
<azonenberg>
then i can try other techniques to flatten it later
<azonenberg>
no go, the cap i dont think is the right approach
<azonenberg>
but yeah still getting some high frequency peaking
<azonenberg>
dont get me wrong, the probe is looking great from dc to 2 GHz, but i think it's letting higher frequencies through too much
<azonenberg>
so looking at resistors only, with my current 200+200+50 i model at 2.4 dB of peaking at 10 GHz vs 50 MHz. With four 100s and a 50, i only get 0.5 dB of peaking
<azonenberg>
But i also modeled 0.2 dB of insertion loss per cm of probe PCB. With a 60mm long probe, i should have 1.2 dB of insertion loss from the PCB, and thus only 1.2 dB of peaking
<azonenberg>
so i *think* if i'm mathing this right, i will get a near-flat response if the resistors have 1.2 dB of peaking
<azonenberg>
200-100-100-50 seems to be 1.4 dB of peaking which is close to that
<azonenberg>
So i'm gonna try reworking this probe with a fourth resistor and see what happens
<azonenberg>
this is a high impedance probe and not a transmission line probe, but it convinces me the peaking is an artifact
<Degi>
The tetris semms to have lower bandwidth
<azonenberg>
Yes it does
<azonenberg>
its a 1.5 GHz probe
<azonenberg>
mine is easily >2
<azonenberg>
also 100-100-100-75-75 is even flatter (0.3 dB peaking) but i can't test that as i don't have any 75s
<Degi>
Hm ideally you'd just use a stack of 75's since they look the best in the datasheet
<azonenberg>
six 75s would work but is mechanically unworkable on the current pcb, at least with an enclosure
<azonenberg>
i can't get in there to solder that far up :p
<azonenberg>
let me sim that though
<Degi>
lol
<Degi>
Like datasheet says Z/Zo is fine to 10 GHz or something
<azonenberg>
Six 75s actually would give 0.23 dB of loss
<azonenberg>
at 10 GHz
<azonenberg>
let me see if i can get better
<azonenberg>
ooh
<azonenberg>
100-75-75-75-75-50 is SUPER flat
<azonenberg>
-0.069 dB rolloff @ 10 GHz
<azonenberg>
I'm gonna buy some more 100s, 75s, and 50s and prototype this
<azonenberg>
my only concern is that it's so long, physically - the attenuator will be 6mm long plus space between the resistors will probably make it more like 10-12mm
<azonenberg>
so beyond a few GHz it will no longer be electrically short
<azonenberg>
i think we'll be ok but i certainly need to test
futarisIRCcloud has joined #scopehal
<Degi>
Better than 75*6?
<Degi>
Ah yes 0.069 dB
<Degi>
Wait negative dB of loss? What did you do
<azonenberg>
no that was S21
<azonenberg>
not loss
<azonenberg>
Lol
<azonenberg>
Six 75s were 0.23 dB
<azonenberg>
if you swap one pair for 100/50 instead of 75/75 you get 0.07
<azonenberg>
going the other way, three 100s and two 75s gives +0.3 peaking
<azonenberg>
so i think four 75s, 100, 50 is the flattest possible with these particular parts
<Degi>
Hm have you tried searching for other resistors with S parameters available yet? Something with less C
<Degi>
Also you could try to design it transmission line like
<Degi>
Between the resistors have a trace of matched impedance, that way it shouldnt be too bad that it isnt electrically short
<Degi>
If you use 6 layer PCB with buried vias, you could have the ground plane like a stair below it, that way the line width shouldnt vary too much
<azonenberg>
No
<azonenberg>
I'm using 2 layer (ish, actually a 4L stackup with signal on 1, ground on 2, and 3/4 unused - just for mechanical reasons) grounded coplanar waveguide
<azonenberg>
the CPW is designed so a 50 ohm trace is 0.5mm wide
<azonenberg>
Which makes it a perfect match to an 0402 footprint
<azonenberg>
But you will still get reflections off the impedance mismatch of the resistor itself
<Degi>
Hm I meant 500 ohm trace on the input, then 400, 315, 250, 175, 100, 50 ohms between the resistors and SMA
<azonenberg>
you mean 500 ohm impedance?
<azonenberg>
that would be ridiculously thin i think
<azonenberg>
And you'd then have to match to the resistor
<azonenberg>
no, the way i have it here with constant width makes sense
<Degi>
I mean the input impedance is 500 and the output 50 ohms, right?
<azonenberg>
Correct
<azonenberg>
Ideally, a transmission line probe is an ideal 450 ohm resistor of zero physical size with perfectly ohmic characteristics and flat frequency response
<azonenberg>
with 50 ohm traces on all sides
<azonenberg>
I think i'm going to be OK with the attenuator being ~1cm long
<azonenberg>
i might have some distortions out in the 8-10 GHz range but honestly if this turns into a 5 GHz probe i'll be over the moon happy
<azonenberg>
the main thing i want to avoid is, if i'm probing a very wideband signal like a fast edge, i don't want massive gain of the upper harmonics like i have now