<azonenberg>
i'm debating pulling back soldermask to cover the stitching vias just for appearances sake
<azonenberg>
but as it stands now all of the fringing fields around the GCPW center trace should be in air or RO4350B, not soldermask
<azonenberg>
which is all that really matters electrically
<azonenberg>
i have soldermask dams on either end of the CPWs around any soldered connections to prevent solder from flowing out onto the traces, and i want mask under the silkscreen text for contrast
<azonenberg>
oh i should add stitching vias along the side where the probe ground is going to make contact too
<azonenberg>
Just sent it off to Betty for a quote
<zigggggy>
generated a wrapper around the xstreamdso com object
<zigggggy>
not 100% sure how it works but i think i have an idea
Famine has quit [Ping timeout: 256 seconds]
Famine has joined #scopehal
<monochroma>
azonenberg: is this a known warning?
<monochroma>
/scopehal-cmake/lib/scopeprotocols/Ethernet100BaseTDecoder.cpp:247:64: warning: ISO C++11 requires at least one argument for the "..." in a variadic macro
<monochroma>
247 | LogTrace("Ethernet100BaseTDecoder: Unable to sync RX LFSR\n");
Famine has quit [Ping timeout: 240 seconds]
<lain>
should just need to define the same macro without variadic iirc
Famine has joined #scopehal
m4ssi has joined #scopehal
_whitelogger has joined #scopehal
m4ssi has quit [Remote host closed the connection]
m4ssi has joined #scopehal
<azonenberg>
monochroma: yes and i have yet to decide what i am going to do about it
<azonenberg>
i bumped up the warning level a bunch lately
<azonenberg>
Bumping the C++ standard to a newer version is probably the best move?
<azonenberg>
or just disabling the warning as ~all compilers allow that even though the standard technically doesnt
<azonenberg>
(C++ 20 allows it officially)
<azonenberg>
Or i could create a new log macro LogTraceNoArgs etc that takes no arguments
<azonenberg>
its not a super high priority but i do want to get rid of it as part of refactoring and warning cleanup
<miek>
i don't think C++20 is well supported across different distros yet
<azonenberg>
Yes, that is why we're sticking with 11 right now
<azonenberg>
What i may do is see if i can just disable the warning since gcc and clang both allow it in c++ 11 mode despite the standard not doing so, and the spam is annoying
<azonenberg>
then when we switch to c++ 20 in ~5 yrs or so when everyone supports it in oldstable distros, deploy a proper fix
<apo>
#ifdef all the things
<azonenberg>
BTW this is a logtools issue not scopehal, as i think logtools's macros are about the only place I use varadic macros in the project
<azonenberg>
But afaik so far scopehal is the only logtools user that is using --std=c++11 -Wpedantic
<_whitenotifier-3>
[scopehal] azonenberg pushed 1 commit to master [+0/-0/±2] https://git.io/JvKlR
<_whitenotifier-3>
[scopehal] azonenberg c170516 - EyeDecoder2: don't segfault if given invalid inputs. Fixes #63.
<_whitenotifier-3>
[scopehal] azonenberg closed issue #63: Find a USB 1.0 Low Speed device to test decoders on - https://git.io/fjZ1v
<_whitenotifier-3>
[scopehal-apps] azonenberg pushed 1 commit to master [+0/-0/±2] https://git.io/JvKlE
<_whitenotifier-3>
[scopehal] azonenberg reopened issue #63: Find a USB 1.0 Low Speed device to test decoders on - https://git.io/fjZ1v
<azonenberg>
gaah another wrong-repo fix
<_whitenotifier-3>
[scopehal-apps] azonenberg closed issue #63: Attempting to add Eye pattern analyzer with NULL clock is a segfault - https://git.io/JvEYu
<azonenberg>
also THIS is interesting
<azonenberg>
i may want to change how i populate the probe characterization board
<azonenberg>
It looks like i will actually get flatter frequency response with a single 50 ohm FC0402 than dual 100s
<azonenberg>
(wrt the 50 ohm load)
<azonenberg>
at 10 GHz a FC0402 50 ohm is 54.6 ohms while parallel 100s are 43.85
<azonenberg>
This translates to return loss of -23.8 vs -13.7 dB
<azonenberg>
Not sure why i thought the dual 100s were better, but good thing i actually looked at the s2p instead of just looking at the (apparently not super accurately rendered) graphs in the datasheet
m4ssi has quit [Remote host closed the connection]
bvernoux has joined #scopehal
<azonenberg>
lain/monochroma: things are proceeding nicely re probe tips from PMK. They forwarded me to their Texas distributor (scientific test inc) who just contacted me to acknowledge that they're drawing up a quote
<azonenberg>
most importantly they're not refusing to deal with me because i'm not a Big Vendor, as i feared might be the case
<bvernoux>
ha great
<azonenberg>
bvernoux: did you see the final iteration of the characterization board?
<azonenberg>
That's not using a thick metal model properly i think
<bvernoux>
to 1mm vs 0.8mm
<azonenberg>
oh wait, that assumes 53um conductor thickness not 35 which i'm using
<azonenberg>
which is a HUGE difference with closely coupled sidewalls
<bvernoux>
I remember that Sonnet have some error about Impedance simulation
<bvernoux>
and SaturnPCB was correct
<bvernoux>
anyway error is something like 2Ohms max
<azonenberg>
We'll see what the pcb fab says
<azonenberg>
Saturn is supposed to be using a polynomial approximation tracking Sonnet results
<azonenberg>
so they should give very similar results
<bvernoux>
there is no impact on copper thickness
<azonenberg>
there is a *huge* difference with and without the thick metal model in a CPW
<azonenberg>
because you get much more coupling in plane
<bvernoux>
ha yes in Sonnet Simulation I remember that
<bvernoux>
but even with thick metal there was something like 1.5 / 2 Ohms error
<bvernoux>
to have very accurate Z it requires tons of cell IIRC
<azonenberg>
I'm using a 2 sheet thick metal model in Sonnet, my 0.8mm CPW is 50.04 ohms @ 100 MHz, rises to a peak of 50.13 ohms @ 1 GHz, drops to 49.8 ohms @ 7.5 GHz, then rises to 49.97 ohms @ 10 GHz
<bvernoux>
Yes
<bvernoux>
but IIRC that simulation is not accurate enough with SOnnet by Default even with Thick Metal
<bvernoux>
for GCPW
<bvernoux>
for Z param
<bvernoux>
let me find the answer from SOnnet
<azonenberg>
I pushed things pretty far. 3992 subsections, 122 MB of ram
<bvernoux>
by Twitter
<azonenberg>
just for a short length of CPW
<azonenberg>
that was as far as i could go in L2 basic
<bvernoux>
but the guy is crazy of perfect things ;)
<azonenberg>
when i see a beautiful wooden box full of micrometer gauges...
<azonenberg>
yeah
<bvernoux>
yes ;)
<bvernoux>
he's using a 100KUSD VNA too ;)
<bvernoux>
with auto calibration which cost > 15KUSD ...
<azonenberg>
I see that. and lol
<bvernoux>
It is why at end my cable with -30dB S11 are not so bad ;)
<bvernoux>
especially so far I cannot go outside 6GHz ;)
<bvernoux>
look the sentence "For me, I would not use a cable that is worse than -40 dB." ;)
<bvernoux>
page 42
<azonenberg>
How does the center conductor solder to the sma pin?
<azonenberg>
Or is that just friction fit?
<bvernoux>
it is friction yes
<bvernoux>
it solder only border/GND
<miek>
oof, those pictures of the bad connectors :(
<bvernoux>
yes
<bvernoux>
that can ruin expensive connectors
<miek>
ooh, those heat sink adapters are cute. might have to make some of those
<azonenberg>
i'm pretty sure the Crystek semi-rigid cables i was using are soldered like that
<miek>
there's another document from that same lab on general VNA usage that's very good, i'll see if i can find a link
<bvernoux>
the good point is with that technic we can build ourself excellent cable
<bvernoux>
which cost less than 10USD per unit
<bvernoux>
such cable cost > 150USD
<bvernoux>
Materials (pricing as of October 2013)AmphenolRF #901-9808 captivated contact SMA connectorsMouser.com $6.07 ea @qty 25Belden #1673A 0.141” o.d. Conformable cable Tallycom.com $4.24/foo
<bvernoux>
need to check ;)
<bvernoux>
price was from 2013
<bvernoux>
I have received some good cable for NFC too ;)
<bvernoux>
I have tested them up to 6GHz
<bvernoux>
and for the price they are excellent and anyway for NFC it is overkill ;)
<azonenberg>
lol
<bvernoux>
I will do not use them for serious RF stuff > 2GHZ anyway ;)
<bvernoux>
but it is perfect for 13.56MHz ;)
<miek>
i do wonder about those gauges too, it seems like they shouldn't be hard to make for a good machinist?
<bvernoux>
such gauges are ultra expensive
<bvernoux>
some are cheap on Ebay but how to be sure they are working correctly ;)
<bvernoux>
like for calibration kit it is never recommended to buy used calibration kit ;)
<azonenberg>
bvernoux: btw speaking of cal
<bvernoux>
as they have probably been used more than 500times and so they are considered as "dead"
<azonenberg>
both of my lecroys are now fully up to date cal wise
<bvernoux>
azonenberg, ha nice
<azonenberg>
And i'm looking into getting my R&S meters and PSUs cal'd properly
<azonenberg>
seeing as they're ~2 years out of date
<bvernoux>
do you have negotiated some price ?
<azonenberg>
well one of the lecroys is new to me, the seller cal'd it for me
<azonenberg>
the other one just got back from service at the factory and i got a free cal out of it
<azonenberg>
for the R&S stuff i'm still waiting on an up to date quote from a local cal place so i dont have to ship them to germany
<azonenberg>
That company does on-site cal work in a mobile temp controlled lab as well
<azonenberg>
i'm considering taking advantage of that next year, rather than trying to ship ~7 pieces of gear out for cal
<bvernoux>
end of calibration date of my VNA is March 2020 ;)
<bvernoux>
But it is well known that HP8753D in same environment with stable temperature does not requires calibration each year like some vendor ask it
<azonenberg>
well one of the first priorities is to HAVE a stable temp :)
<bvernoux>
I have all measurements done and it was very nice for such old HW
<azonenberg>
Which i currently do not. But HVAC equipment is being installed in 3 weeks
<bvernoux>
I have air conditioning with laser temperature sensor with 0.1°C accuracy ;)
<bvernoux>
installation was finished 1 week ago
<bvernoux>
and the room/lab is quite small (and closed) about 12square meter
<azonenberg>
My lab is 37 m^2 and currently has a door open to the house all the time to provide some airflow to dump waste heat
<azonenberg>
in addition to an exhaust fan sucking hot air out to the ceiling
<azonenberg>
to the roof*
<azonenberg>
both will be closed by default once i have hvac in there
<bvernoux>
ha nice
<bvernoux>
Does anyone have bought the KC908 during crowdfounding?
<bvernoux>
It seems i'm the only one outside china ;)
<azonenberg>
that sounds like a kintex devboard name
<bvernoux>
yes
<azonenberg>
except it's not
<bvernoux>
it is based on AD9361 with lot of custom things
<bvernoux>
MCU + FPGA ...
<bvernoux>
and some nice Preselection filters + ext AMP ...
<bvernoux>
the blog of the developer is interesting
<miek>
so that gauge is really just a standard dial indicator + an adapted head on it to measure the pin depth / dielectric depth + some standards to compare against. get a nice indicator off the shelf (new mitutoyo ones are £35 for 10um resolution, £85 for 1um), build the adapter (this does not need to be accurate), and build some standards (need a surface plate, height gauge, and good gauge blocks to inspect).
<miek>
"easy"!
<bvernoux>
it explains all the mess to reach a noise floor of less than -120dBm with the AD9361
<bvernoux>
miek, will be interesting to have such gauge for < 50USD ;)
<bvernoux>
in the blog there is some PCB described too
<bvernoux>
but the project is not open source :(
<bvernoux>
Anyway the guy is quite open to speak about the architecture with lot of details
<bvernoux>
He do not want to provide firmware as he say no one will do anything with the code as it is too complex
<bvernoux>
which is not really wrong in 99% of case
<bvernoux>
a good example is airspy firmware it is fully open source but there was pratically no contribution
<bvernoux>
and it is 1000x time less complex than the KC908 ...
<miek>
i don't know if that's the best example :) that project/company's got an.. interesting.. relationship with open source
<bvernoux>
also hackrf is a good example ;)
<bvernoux>
I worked on that project since the start
<bvernoux>
and we can say major contribution on HW are 0
<miek>
i think hackrf's had a lot of good contribution!
<bvernoux>
as it requires some expert in the domain with time and motivation (and HW)
<bvernoux>
I speak about contribution related to HW
<bvernoux>
to remove some spurs and quirks
<bvernoux>
mainly for example changing SI5351 voltage to 1.8V
<bvernoux>
it improve lot of things
<bvernoux>
I have done that on AirSpy
<bvernoux>
with lot of tests ...
<bvernoux>
lot of optimisations are possible but it is quite time consuming ...
<bvernoux>
instead of that there was some clone without any improvement ...
<bvernoux>
miek, on major contributors to HackRF I count about 8 (and you are in ;)
<bvernoux>
I count those with >= 16 commits
<miek>
yeah, collaboration on HW is difficult anyway. getting volunteers to get stuff upstream is unlikely, but it's not all about improvments - i think having it all open is great for the users
<miek>
and there was the rad1o badge, which was a pretty awesome thing to come out of it
<bvernoux>
yes it is amazing to have such RF stuff open HW
<bvernoux>
yes rad1o badge is the most ambitious
<miek>
thanks :) yeah, the volume of FW contributions may not be huge, but i think some really neat things have come out of it
<bvernoux>
the main issue is to have more and more RF guys which want to work in Open HW stuff ;)
<bvernoux>
as HW are still ultra expensive
<bvernoux>
even today 1GHZ BW Scope is still too expensive
<bvernoux>
I see also price on Ebay on old HW which is very expensive
<bvernoux>
HP Agilent 8753D 30 kHz to 6 GHz => > 5KUSD
<bvernoux>
I have paid mine < 2500USD with Full Calibration & Warranty
<bvernoux>
It's crazy
<bvernoux>
On the other side it is nice to see some ultra cheap VNA like nanoVNA for less than 70USD ;)
<miek>
yeah, i really like that this sort of thing is becoming available. there are so many cases where you don't need the top-end calibrated gear, a rougher measurement will do
<bvernoux>
yes it is amazing to see such price for that quality
<bvernoux>
the NanoVNA 2 is announce to do up to 3GHz ;)
<bvernoux>
for < 100USD ;)
<bvernoux>
could be a game changer for lot of guys
<azonenberg>
bvernoux: yeah i am planning to build a f/oss scopehal compatible DSO in the not too distant future based on the HMCAD1520
<bvernoux>
ha yes will be great
<azonenberg>
with various configurations ranging from 160 Msps 12 bit @ 4 channels up to 1 Gsps 8 bit @ 1 channel
<bvernoux>
amazing as today fast DSO does not exist
<bvernoux>
or are clearly too expensive
<azonenberg>
Yeah this is the same adc rigol uses in some of their lower end gear
<bvernoux>
what is the estimated price point ?
<bvernoux>
DSO+TDR ;)
<azonenberg>
The HMCAD1520 is $113 on digikey @ qty 1
<azonenberg>
The pin compatible HMCAD1511 is $64 and lacks the HD mode (8 bit only)
<bvernoux>
it seems it is replaced by ADS5401
<bvernoux>
TI recommend to switch on this one ;)
<bvernoux>
12bit 800MSPS
<azonenberg>
Thats a single channel, the HMC has four inputs and muxing etc
<bvernoux>
166.76USD / 1ku ...
<azonenberg>
i'm talking about a one adc 4 channel scope
<bvernoux>
ha ok
<azonenberg>
cost optimized
<bvernoux>
yes
<bvernoux>
ha yes it's an AD chipset not TI ;)
<azonenberg>
it's a hittite microwave chipset
<azonenberg>
that AD bought
<azonenberg>
anyway, pair it with say an XC7S25 ($28.91) or XC7S50 ($48.30)
<bvernoux>
especially if it is done in the ADC I suspect it is embedded decimation ...
<bvernoux>
will be amazing on 4Layers
<bvernoux>
to limite the cost
<bvernoux>
for signal integrity it will be probably hard ...
<bvernoux>
as I doubt you can keep some layers reserved for GND & PWR Supply
<azonenberg>
block ram only, XC7S50 has 150x 2KB block RAM so ~300 K points max assuming no bram used for anything else, probably 200K points after some buffers for other stuff
<azonenberg>
so 200K points on 1ch / 100k on 2/ 50k on 4
<azonenberg>
Which is probably fine for an entry level scope
<bvernoux>
yes
<azonenberg>
I've done 7 series designs on 4L with S-G-P-S stackup in the past
<azonenberg>
it should be doable
<azonenberg>
one internal layer for all power, one for solid ground
<bvernoux>
I'm afraid of cross talk and other issue with such density on 4L
<bvernoux>
especially for 4Chan
<azonenberg>
i dont think it will be that bad, but we'll see. The AFE is a big TODO
<azonenberg>
i basically know how the rest of the system will be architected
<azonenberg>
btw my plan is to be 50 ohm input only, no high-Z
<bvernoux>
yes
<azonenberg>
i'll make a low bandwidth active probe at some point to pair with it
<bvernoux>
50 ohm is perfect
<azonenberg>
for when you need high Z
<bvernoux>
for small signal up to 5V max ...
<azonenberg>
but default is for transmission line probes
<bvernoux>
will be amazing with scopehal ;)
<bvernoux>
with a decent PC
<azonenberg>
well thats the fun bit. Yes, only 50k points on all channels
<bvernoux>
with USB 3.0 or Eth GigaBit
<bvernoux>
I think Eth Gigabit is better for streaming
<bvernoux>
multi Gigabit
<azonenberg>
BUT... 50k points * 8 bit * 4 chans = 1600 Kbits = 1.6 Mbits per waveform
<azonenberg>
so theoretical limit of 625 WFM/s @ max depth capture over gig-e, not counting framing overhead
<azonenberg>
with overhead, i think 500 WFM/s will be doable
<bvernoux>
yes will requires 10Gb ETH ;)
<azonenberg>
No, over 1G
<bvernoux>
ha yes
<azonenberg>
With 10G you'd be able to get more like 5 kWFM/s
<azonenberg>
also this is 500 WFM/s at max memory depth. Less deep would let you go further
<azonenberg>
Tentative thought for initial firmware is to use UDP and simply consider a bit error a missed trigger
<azonenberg>
i.e. if you fail to receive valid data for all waveforms in a capture, discard it and pretend the scope never triggered
<azonenberg>
the BER on a LAN should be low enough you can get away with that
<azonenberg>
although i'd like to move to TCP at some point
<azonenberg>
but to start it will be a simple unauthenticated UDP control plane channel on one port, plus a pub/sub interface
<azonenberg>
once activated the scope will just spam waveforms to the currently subscribed client (if any) every time it triggers
<azonenberg>
obviously not something we'd keep forever, but would be good enough for a PoC i think
<bvernoux>
or using QUIC ;)
<azonenberg>
I already have a FPGA UDP stack so that would be quickest
<bvernoux>
yes
<bvernoux>
for test UDP is perfect
<azonenberg>
anyway, so hardware project status... probe characterization board is waiting for fabs to turn around quotes
<azonenberg>
probe needs a respin for the new probe tip socket but that's on hold until the new characterization board comes back so i can do final testing on the current hardware rev
<bvernoux>
for me the most interesting is DSO with Eye Diagram on 10GHz signal ;)
<azonenberg>
clock distribution system is in the early stages of schematic capture, i haven't touched it in a while. I want to get back to that soonish
<bvernoux>
yes clock distribution is a must have too
<azonenberg>
i think what might make sense to do soonish is a standalone AFE for the low speed scope, just to test/characterize
<azonenberg>
SMA probe input, 50 ohm buffered SMA output to the ADC (or a DSO/VNA during characterization)
<bvernoux>
yes step by step
<azonenberg>
then a little stm32 with a 3.3v uart for a serial console to configure the gain/offset etc
<azonenberg>
on oshpark board
<azonenberg>
For 1 Gsps scope i figure 250 MHz B/W on the AFE should be plenty?
<azonenberg>
also we may want to consider switchable antialiasing filters since with 4 chans the nyquist b/w will be 125 MHz
<bvernoux>
yes 250MHZ BW is really the max for 1GSPS
<azonenberg>
but i feel like putting a 125 MHz filter on a single 1 Gsps channel would be a waste
<bvernoux>
I always count /8 to have something usable
<bvernoux>
so 125MHz
<bvernoux>
it depends on signal anyway
<azonenberg>
yes but the scope will have multiple modes depending on bit depth and channel config
<azonenberg>
my question is, do we want reconfigurable antialiasing filters?
<azonenberg>
Or do we want to pick one and call it good?
<azonenberg>
Since it's an entry level device i dont want to add a ton of complexity to the afe
<bvernoux>
yes
<bvernoux>
it is a good question
<azonenberg>
maybe just put a 100 MHz filter and stick with that?
<azonenberg>
In quad channel 160 Msps mode you may get a little bit of aliasing since nyquist is 80 MHz, but it will be enough for everything else
<bvernoux>
160MSPS mode is with 14bits?
<azonenberg>
12 bit 4 ch
<bvernoux>
ok
<azonenberg>
The total set of configurations i see in the datasheet are...
<azonenberg>
Having RF relays and multiple filters, or something like that, is not beyond the realm of possibility but i do want to keep costs reasonable
<bvernoux>
yes
<azonenberg>
For the entry level scope i'd rather remove features and keep it affordable
<azonenberg>
But have the features we keep work reliably
<electronic_eel>
aren't the filters in AFEs usually switched with some solid state device?
<electronic_eel>
the rf relay is usually for 1M and 50R
<azonenberg>
I dont plan to have 1M probe support at all
<azonenberg>
I consider them deprecated :p
<electronic_eel>
I know - so no rf relay
<azonenberg>
I think i can make a nice active probe for less than the crazy marked up cost of an 1M ohm OEM probe
<azonenberg>
at least for low bandwidth. And transmission line probes past that
<azonenberg>
and i'm not sure about the best way to do the reconfigurable filter
<azonenberg>
Ideally i guess we'd want a filter a bit below nyquist for each of those configurations
<electronic_eel>
switch capacitors in the feedback path with some fets?
<electronic_eel>
or switch between different feedback paths with a mux or diodes
<azonenberg>
Yeah. I'll have to start doing some research
<azonenberg>
I guess that might be tonight's project lol
<electronic_eel>
but depends on how the AFE is designed
<azonenberg>
Yeah. So basically the front end will consist of a 50 ohm termination (probably just a normal resistor, that should be flat enough out to the few hundred MHz this AFE is meant for)
<azonenberg>
then some kind of ~unity gain amplifier with a DAC on the other input to provide offset control
<bvernoux>
for multiple filter the only way is to use RF switch
<azonenberg>
then a variable gain stage
<azonenberg>
after that, the bank of antialiasing filters
<bvernoux>
or even cascade some RF switch to have better isolation ...
<bvernoux>
RF switch are not very expensive but depending on how many there is signal shall be amplified ...
<electronic_eel>
don't forget input protection
<bvernoux>
and filtered ...
<bvernoux>
at the end it is a quite complex preselector ;)
<azonenberg>
electronic_eel: So, what sort of protections? ESD clamp diodes for sure
<azonenberg>
Do we plan to have anything past that? Any protections against a hard DC input beyond the specced input range?
<bvernoux>
yes mixing 2 types of ESD diode is very efficient
<bvernoux>
and potentially adding a good DC block
<azonenberg>
So switchable 50 ohm ac/dc input?
<electronic_eel>
yes, but if you use basic tvs diodes they add tons of capacitance
<bvernoux>
but i doubt you want DC block especially to have low freq ...
<azonenberg>
I was thinking dc only for a low speed scope likethis
<azonenberg>
and just have a moderately wide offset range
<electronic_eel>
but as bvernoux said, you mix several types of diodes to get low capacitance and good performance
<bvernoux>
some TVS diodes have low capacitance < 0.5pF
<azonenberg>
electronic_eel: The ESD133, for example, is 0.2 pF @ 1 MHz
<bvernoux>
it is the price to pay for ESD protections ...
<azonenberg>
I may also add a series resistor to limit input current. Given that this is a stupidly low frequency design, i think we can afford to be quite aggressive on the protections
<electronic_eel>
the ESD133 are very small. they will work well for small esd transients - but you will easily fry them when you apply a too high dc voltage
<azonenberg>
electronic_eel: i guess the other question is, how idiotproof do we want to make the design?
<azonenberg>
if we assume the scope will mostly be used with 10:1 probes, you'd need 50V to go over a 5V Vimax
<azonenberg>
esd protection from transients during handling, mating/unmating probes, etc makes sense
<azonenberg>
but it would take effort to fry the input
<azonenberg>
if doing a direct SMA connection we'd assume the user has some idea of what they're doing
<electronic_eel>
is 10:1 really the only scenario? what about a user directly connecting a signal source or function gen?
<bvernoux>
yes anyway user shall be ESD safe like for any T&M stuff
<electronic_eel>
then you can easily go over 5V
<electronic_eel>
I have done some research for a input protection addon board for Glasgow
<bvernoux>
adding 2 ESD diode to protect correctly the device work fine up to more than 1GHz
<electronic_eel>
you mean text or a special test I describe there?
<azonenberg>
I'll look into that more tonight after work. Looks like some good info though
<electronic_eel>
I received some more equipment and got some further ideas regarding the esd protection part since I wrote it
<azonenberg>
anyway for the initial characterization board i will probably not have any overvoltage protection, just some clamp diodes
<electronic_eel>
so I'll probably add some of that next week or so
<bvernoux>
electronic_eel, the comparison of different protection ...
<electronic_eel>
yes, if you build it in a modular way you can always plug the protection in front of the input
<azonenberg>
Yeah exactly
<bvernoux>
it will be interesting to check with a VNA ;)
<azonenberg>
My vision for the higher end scopes is actually to have the AFEs, ADCs, and FPGAs on separate boards to the extent practical
<bvernoux>
like always SA are not enough in lot of cases
<azonenberg>
Such that you can blow out an input and replace only the damaged board
<azonenberg>
and, more importantly, during development if the AFE board doesn't meet spec, you don't have to throw out a $3000 ADC
<electronic_eel>
bvernoux: ah yes. that part of testing is still missing. I'm not too familiar with using a vna so will have to do some learning there
<azonenberg>
i will be characterizing this AFE with my xaVNA
<azonenberg>
but it doesn't have great low frequency performance
<azonenberg>
i think it officially doesnt go below like 130 MHz
<bvernoux>
electronic_eel, especially on the version when you add 2 transistor
<bvernoux>
testing that with a VNA from 30KHz to 2GHz shall show some interesting things
<electronic_eel>
I have a VNWA3, but haven't done much with it yet. but it works well for low freq stuff
<bvernoux>
yes clearly for low freq it shall be good what I suspect is the effect on high freq
<bvernoux>
I confirm BAV99 are very good ;)
<bvernoux>
I have not found something better up to 1.8GHz for good protection against ESD ...
<bvernoux>
for SDR purpose
<azonenberg>
yeah i want from ~dc out to ~100 MHz
<azonenberg>
as that's the range for the entry level scope
<electronic_eel>
the circuit is designed for the digital IOs of Glasgow, the FPGA can do 150MHz max. So testing to 1 GHz should be enough for the harmonics
<bvernoux>
electronic_eel, anyway your test with Siglent+TG is good as this SA is very nice for the price I will say the best quality price of all SA
<miek>
i can test stuff down to 300kHz on my agilent if it's any help
<azonenberg>
miek: once i have a couple of prototypes of the AFE + protection I'll send them around to whoever has appropriate instrumentation
<electronic_eel>
bvernoux: yes, the Siglent SA works well, especially since you only pay the lower freq version and upgrade it for free
<bvernoux>
yes it is why the price is very nice ;)
<azonenberg>
The planned AFE will be SMA input and output, a barrel jack or two for power input, and a 3.3V UART for config
<electronic_eel>
miek: my VNWA3 should be usable down to 1kHz
<bvernoux>
electronic_eel, you shall add LXES15AAA1-133
<bvernoux>
;)
<bvernoux>
It is amazing for RF too
<bvernoux>
electronic_eel, I use it in addition to BAV99 with very good result
<bvernoux>
electronic_eel, they are complementary
<miek>
bvernoux: bad news, apparently it's to be discontinued
<bvernoux>
miek, LXES15AAA1-133
<bvernoux>
?
<miek>
yup
<bvernoux>
I do not see it is EOL
<bvernoux>
on DigiKey/Mouser
<miek>
it's not recommended for new designs on digikey here
<bvernoux>
ha yes
<bvernoux>
but so far I have not find something better which replace it
<electronic_eel>
thanks for the recommendation. will add some to my next Mouser order and will test them
<azonenberg>
silver is actually a little less expensive than enig and has lower loss at microwave frequencies, so i plan to use it on any of my boards with unmasked GCPWs
<bvernoux>
ha yes very strange that impact on ENIG
<azonenberg>
basically you start getting problems when your skin depth is on the order of the nickel plating thickness
<azonenberg>
almost all of your signal is in the plating, not the copper
<bvernoux>
especially on GCPW
<azonenberg>
Silver, on the other hand, is actually more conductive than copper
<azonenberg>
but not by a lot. Overall loss is ~indistinguishable from unplated copper
<bvernoux>
I understand now why there is bare copper option as it is the best for lower loss
<azonenberg>
yeah, until it corrodes
<bvernoux>
yes ;)
<azonenberg>
Which is why so much RF stuff uses silver
<bvernoux>
yes Immersion Silver is clearly a must
<azonenberg>
I'm not modeling plating in any of my analysis because silver and copper are fairly close in resistivity, assuming the conductor is copper will give a pessimistic estimate of loss
<azonenberg>
But that was a conscious decision supported by data, not laziness :)
<bvernoux>
hehe
<bvernoux>
Immersion Silver has some drawback too
<bvernoux>
Disadvantages of Immersion Silver
<bvernoux>
Low resistance to handling and contamination. Gloves must be worn when handling unsoldered surfaces.
<bvernoux>
Shorter shelf life compared to HASL and ENIG.
<bvernoux>
It reacts to sulphur and will result in degradation of the circuit board when exposed. It is not suitable for application where the board may be exposed to air pollution or environment where sulphur is present.
<bvernoux>
Contact resistance is higher thus it may cause issues during ICT (In-Circuit Test).
<bvernoux>
be
bvernoux has quit [Quit: Leaving]
<electronic_eel>
but it helps against rf vampires ;)
<apo>
can't write bloodbath without dB
<azonenberg>
from people i've talked to, in most cases the sulfur tarnish is not a huge issue
* zigggggy
packs his bags and moves in with azonenberg
<zigggggy>
azonenberg said he's lives in the boonies
<azonenberg>
zigggggy: well, i'm also a fairly short distance from seattle which was one of the first/biggest outbreaks in the US :p
<monochroma>
:D
<azonenberg>
And it seems to have jumped the water, as the island now has two cases too. None in my city yet but i expect it's only a matter of time
<zigggggy>
azonenberg but do you have enough tp?
<azonenberg>
the local walmart's shelves are bare
<azonenberg>
We have a dozen-odd rolls and just order a giant case on amazon because i expect the panic buying to last longer than those few rolls will
<zigggggy>
lol
<electronic_eel>
why the hell is tp such a issue? if worse comes to worse you can use old newpapers or something
<azonenberg>
electronic_eel: i think people are stocking up in the expectation of being afraid to leave the house for a month
<zigggggy>
because it starts with a few people panic buying tp to clear shelves because stores base inventory on normal buying habits
<zigggggy>
so other people get nervous and also start buying
<azonenberg>
yeah i calculated ~5% of my city buying tp in a week would exhaust walmart's stock
<zigggggy>
then more and more and the media picks up on it
<azonenberg>
5% of the households*
<zigggggy>
yeah.. store real estate is expensive.. everyone does just in time
<zigggggy>
no reason to have huge stock of anything
<electronic_eel>
azonenberg: I fully understand that people plan to not leave the house. But I'd first stock on important food and medical supplies you may need
<azonenberg>
Yeah agreed
<zigggggy>
just give me the virus already so i dont have to worry every time i go out
<azonenberg>
lol
<zigggggy>
let me be done with it
<zigggggy>
amirite azonenberg?
<electronic_eel>
i won't die if I my favorite chocolate for a month
<electronic_eel>
*miss
<azonenberg>
electronic_eel: yeah, the panic buying is annoynig when it stops me from buying stuff i normally use though
<electronic_eel>
yeah, that is really annoying.
<azonenberg>
e.g. IPA and gloves are sold out in a lot of suppliers. Luckily I have a reasonable stash in my lab
<zigggggy>
azonenberg does your lab have hand sanitizer?
<electronic_eel>
luckily I use the IPA for cleaning my pcbs, so I have a big canister of it
<azonenberg>
zigggggy: no, but i have plenty of soap. Which works just as well
<zigggggy>
azonenberg only if you have plenty of water
<azonenberg>
If i really needed to make some sanitizer i have glycerin, carbomer 940, two big bottles of everclear, triethanolamine, and a stick blender
<azonenberg>
(everclear is a very nice source of non-denatured technical grade EtOH for solvent purposes)
<azonenberg>
zigggggy: i'm not expecting water to go away any time soon
<azonenberg>
this is a lung virus, not a zombie apocalypse
<zigggggy>
what if you are not near water
<azonenberg>
in a lab? that's not a super common situation
<zigggggy>
like say.. YOUR car
<zigggggy>
or at the store
<azonenberg>
That's where the gallon jug of sanitizer i picked up years ago for use as a fire starter might come in handy
<azonenberg>
but so far we havent even run out of the little bottle on my wife's purse
<azonenberg>
electronic_eel: i use IPA for pcb cleaning too but i use tiny bits of it on swabs
<azonenberg>
to give you an idea, i recently threw out the dregs of a bottle i had since i was probably an undergraduate
<azonenberg>
that tested at >400 ppm peroxide
<awygle>
yeah i ran out of IPA and this was a bad time for that to happen lol
<azonenberg>
do you know how long ipa has to sit around in an open bottle to oxidize that badly?
<azonenberg>
i have one smallish (16 oz or so) bottle of acs reagent grade that will probably last me the next five years of lab cleaning purposes
<azonenberg>
i use acetone for cleaning decapped chips so that gets consumed in higher volume
<azonenberg>
If i need alcohol for cleaning stuff i'm just gonna use everclear
<azonenberg>
for disinfecting*
<azonenberg>
ethanol works just as well as IPA for that
<azonenberg>
in fact, most sanitizers are EtOH based not iPrOH
<azonenberg>
i havent checked the liquor aisle in the store lately but i'm willing to be there's plenty of such stuff to be found
<azonenberg>
while everyone is panic buying isopropyl
<electronic_eel>
buy a good wine, savor it and relax - best medicine while there is panic outside...
<azonenberg>
lol
<azonenberg>
well i'm under orders to work from home until end of month at which point they'll re-evaluate
<azonenberg>
As someone who lives two hours from the office and hates losing 4 hours out of his day to the commute, this is just fine by me :p
<azonenberg>
that's 4 hours a day i can spend on scopehal stuff etc instead
<electronic_eel>
:)
<awygle>
i am under similar orders
<azonenberg>
i mean i WFH a lot anyway, but 100% remote is just dandy
<awygle>
which makes a lot of sense to me as we've had double-digit deaths within 5m of here
<electronic_eel>
awygle: where are you located?
<azonenberg>
you have 10+ dead people within five meters of you?
<awygle>
redmond
<awygle>
miles, or minutes
<azonenberg>
Oh, lol
<awygle>
meters would be... troubling
<azonenberg>
yeeah
<awygle>
but yeah that place in kirkland is like... not far from here at all
<azonenberg>
i mean i've been that close to ONE dead body at a time. 10+ would be a different experience :p
<azonenberg>
electronic_eel: a fair number of the folks in this channel are in/very close to seattle
<azonenberg>
at least five off the top of my head
<awygle>
huh i must be missing one then
<azonenberg>
awygle: me, you, catgirls, error_404
<electronic_eel>
I'm from Germany. there were like 5 confirmed cases here in the town, they were visiting italy, but thats about it
<azonenberg>
electronic_eel: there's a nursing home near awygle that had 35 tests done on residents recently
<awygle>
ah, error was who i'm missing (assuming i'm parsing "catgirls" correctly which i think i am)
<azonenberg>
33 positive, one negative, rest inconclusive
<azonenberg>
sorry 31 positive
<azonenberg>
that's in one building
<electronic_eel>
wow
<awygle>
i went to my doctor for a routine checkup like two weeks ago, and he was like "yeah it's here already and tons of people have it and don't know it, everyone in the medical community knows it, and we're all probably going to get it too."
<azonenberg>
So yeah it's starting to get bad here. Me and the catgirls are across the sound from all that, but there are now two confirmed cases this side of the water
<awygle>
like "thanks doc but can you refill my prescription or....?"
<electronic_eel>
bit of a contrast to the video where I saw Trump telling "everything is under control, sealed nearly airtight"
<azonenberg>
awygle: yeah IMO the goal of all of this isolation isn't necessarily to *not* get sick
<azonenberg>
the goal is to slow down the rate of spread, and delay when you get sick, to the point that the healthcare system isn't overwhelmed
<awygle>
that graph with "hospital carrying capacity" on it really clarified things for me
<azonenberg>
if 50% of the population gets it all at once, we're screwed. 50% a few at a time over months, much less so
<awygle>
hadn't thought of it in 100% those terms before
<azonenberg>
and if you're lucky and end up in the 50% who dont get it ever, great
<azonenberg>
but if you DO get it, you want it to not be when the hospitals are all stuffed to the gills and half their staff are sick too :p
<electronic_eel>
so get it now when the hospitals aren't filled yet
<azonenberg>
lol
<azonenberg>
Given how much i get out of the house, i think i'll take my chances sitting tight for a while :p
<electronic_eel>
but as incubation takes like a week, you might already be late
<azonenberg>
Can't get sick if you're never around other people :p
<awygle>
yeah i am depressingly certain i already have it tbh
<electronic_eel>
yeah, when you are working from home and don't need to go shopping, there is a good chance to not get it
<azonenberg>
yeah i made one walmart trip at off hours (less people around to sneeze on me) this week to stock up
maartenBE has quit [Ping timeout: 240 seconds]
<azonenberg>
the only other time i've left the house was for SAR training which mostly consisted of sitting out in the woods alone under a tarp for a few hours, then driving a SUV with nobody else in it around running radio relays
<awygle>
sore throat. and i had a colleague whose whole family was out with "flu-like symptoms" for like a week before we got told to stay home
<monochroma>
awygle: there is also a pretty nasty flu going around in the seattle area in parallel with the virus as well from what i have heard
<awygle>
yeah, it's complex
<azonenberg>
yes, we had like three or four people from $dayjob catch that flu
<awygle>
and honestly the treatment either way is "stay home, try not to die"
<azonenberg>
at least one was tested for coronavirus and came up negative
<azonenberg>
that was just before we got told to stay home too
<awygle>
"if you think you might die, _then_ go to the hospital"
<electronic_eel>
wasn't the coronavirus test used in the us until a few days ago faulty?
maartenBE has joined #scopehal
<azonenberg>
electronic_eel: UW has their own test
<awygle>
yep. UW developed their own, which is "unofficial"
<azonenberg>
developed by the virology crew at UW med school
* monochroma
wonders when the looting is going to begin when everyone figures out all the businesses are empty... v_v
<azonenberg>
But if you caught a mild case of it, a) stay home and don't infect anyone else, and b) you probably won't get a bad case later on
<awygle>
absolute clown show tbh
<awygle>
just gonna keep drinking two gallons of tap water a day and sleeping a lot