electronic_eel has quit [Ping timeout: 246 seconds]
electronic_eel has joined #scopehal
m4ssi has joined #scopehal
futarisIRCcloud has quit [Quit: Connection closed for inactivity]
m4ssi has quit [Remote host closed the connection]
m4ssi has joined #scopehal
<azonenberg>
So these resistors are actually quite a bit better controlled than the datasheet would have you think, lol
<lain>
probably per-batch is very tight
<lain>
but across batch is looser
<azonenberg>
They're nominally +/- 0.1%
<azonenberg>
So with 450 ohms that would give you +/- 0.45 ohms or 449.55 to 450.45 as the value of the resistor string
<azonenberg>
biased toward the center assuming the distribution of each of the six resistors is independent
<azonenberg>
You know what the actual resistance values i measure across the entire probe (tip to SMA, including solder joints and pcb traces?
<azonenberg>
450.23 to 450.39 ohms, average of 450.31, and a standard deviation of 44 MILLIOHMS
<azonenberg>
That's +/- 80 milliohms of probe to probe variation, or +/- 0.017%
<azonenberg>
That is *amazingly* consistent
<azonenberg>
And that includes solder joint and pcb copper tolerances, not just the resistors
<azonenberg>
although i imagine most is in the resistor
<azonenberg>
Assuming the resistors themselves have a distribution centered at 450.00, that would mean the SMA + socket + trace + solder joints have around 30 milliohms of total resistance, which sounds reasonable
<azonenberg>
(According to some google calculators (didnt bother with more accurate modeling) ~65mm of 0.31mm wide trace in 1 oz copper should actually be 100 milliohms, i wonder if the plating is reducing that?)
<azonenberg>
all of the probes from the kickstarter batch are assembled. One had a tip socket damaged during testing and needs to be replaced, i'm characterizing all of the others
<azonenberg>
here's the data i've collected so far
<azonenberg>
I'm doing the full professional edition workup on everything, even the standard edition ones for my backers, just to collect additional data on variability from unit to unit
<azonenberg>
we were just commenting on how the nominal +/- 0.1% resistor string seems to have actual variability much less than that
<azonenberg>
the overall DC resistance from tip socket to SMA is +/- 0.017% of the average measured value
juli964 has joined #scopehal
<electronic_eel>
nice spreadsheet
<electronic_eel>
is there any visual difference between s/n 0019 and the others?
<electronic_eel>
it has a bit lower bw and higher risetime than the others, but the dcr is about the same as the others
<azonenberg>
I was going to go back and visually inspect post characterization of everything
<azonenberg>
if it's the soldering at the SMA, i can't see - they're already glued into enclosures
<azonenberg>
but if it's the tip or resistor string i can see it
<azonenberg>
i did observe that as well and it was interesting
<azonenberg>
The numbers jump around a bit because i made 25 labels, printed 20 enclosures
<azonenberg>
assigned them last to first because thats how the labels were printed
<azonenberg>
then grabbed them out of the bin randomly to build 15 units
<azonenberg>
I have full scope waveforms of the rising edge for each probe and full 2-port s-parameters, there's a lot more data than is in the sheet
<azonenberg>
i'll be uploading raw data to starshipraider-caldata in an bit
<azonenberg>
i also have to still rework s/n 0017 and collect input impedance measurements (not across a terminator)
<miek>
it's an electronic calibration module. rather than connecting SOLT standards to all your vna ports, you connect an ecal and it switches through them all automatically
<Degi>
Neat
<Degi>
And that costs 10k?
<azonenberg>
whats the main purpose, avoiding connector wear?
<azonenberg>
i want one :p
<azonenberg>
(i've known they existed for a while, just couldnt justify the budget lol
<miek>
i think the main purpose is probably just making it easier/faster, but less wear is an advantage too
<azonenberg>
but these probes are doing better than i thought, so i am planning to tighten up the spec range a bit so the min at -23.5 dB (3 dB down from the nominal -20.5) is actually something like 2.25 GHz
<Degi>
Not 1.9?
<Degi>
Ah nominal is -20.5
<azonenberg>
Yeah
<azonenberg>
Every other probe will pass that spec
<azonenberg>
you can see in the second screenshot just how much 0018 is underperforming
<azonenberg>
and then 0020 is an awesome grade-A probe top blue trace)
<azonenberg>
the rest are all pretty repeatable
Nero_ has joined #scopehal
Nero_ is now known as NeroTHz
<miek>
azonenberg: btw, did you take a look at that EECS142 link from the other day? lab #1 & the video with Joel may be very interesting to you - it's all about measuring & modelling passive component parasitics
<azonenberg>
halfway, meaning there are problems with the license file
<Degi>
awh
<azonenberg>
i have the full solver but the editor isn't letting me use all the features
<azonenberg>
effectively, until they straighten it out i can't use conformal mesh, >2 layers, >8 ports, s-parameter components, etc since those limitations are enforced in the editor
<azonenberg>
but i can use the full 32-core solver with no cap on design size within those 2 layers
<electronic_eel>
do you think it should work or does the limitation look intentional?
<azonenberg>
It should work. I'm going back and forth with support now
<azonenberg>
half the components still think they're on l2 basic, the other half think i'm on pro
<azonenberg>
i think the license file was configured wrong
<azonenberg>
Anyway I have a sim running right now that is a simplified version of the full AKL-PT1 probe with the resistors shorted out
<azonenberg>
it's using 5GB of ram and is maxing out my workstation pretty nicely :p
<Degi>
Neat
<azonenberg>
so even gold would not be enough for this, however it's close enough to feasible that if i split it up into ~3 subsections, which should be doable, it would work
<azonenberg>
and gold would let me model the actual resistor s-parameters
<azonenberg>
Just a question of if i can come up with $8K ish before the end of the year as that's what it will cost me to upgrade
<azonenberg>
and my support contract is up at the end of the year
<azonenberg>
so my options are to do nothing and continue using basic but not have any upgrades or tech support
<azonenberg>
pay a few hundred bucks to renew my support contract for basic
<azonenberg>
pay around 4k to upgrade to silver, or pay around 8k to upgrade to gold
<Degi>
Pay 4 k to a software dev to make OpenEMS usable?
<azonenberg>
it would take a lot more than that
<azonenberg>
i did the math already
<azonenberg>
i would love openems to be usable but it's far enough from complete now that i'm much better off paying for a sonnet seat than hiring a dev to fix openems
<Degi>
awwwh
<azonenberg>
and i do not have the budget to pay for the amount of work openems actually needs to do what i want
<azonenberg>
it would be IMO a kicad+cern level project
<azonenberg>
major sopnsorship with multiple full time devs for years
<Degi>
Hmh, I wonder how Sn96.5/Ag3.0/Cu0.5 comparess to the Sn96.3/Ag0.7/Cu03 of SAC305
<azonenberg>
Hmm i'm not sure why you'd lower the silver content, cost? i've seen 3% Ag / 0.1% Cu which is supposed to be better for certain mechanical properties
<azonenberg>
but 305 seems to be the industry standard right now
<Degi>
Hmh mouser doesnt seem to have 305 from chipquik or at a reasonable price, I think the Ag3.0 should be fine...