<azonenberg>
Initial impressions: second round footprint for the Amphenol SMA has too *large* mounting holes
<azonenberg>
it fits but has too much slop
<azonenberg>
the connector moved during reflow and is now misaligned
<azonenberg>
The probe tip fit is really tight
<azonenberg>
I need to leave a bit more length in the legs
<azonenberg>
Welp
<azonenberg>
I'm actually quite happy with how it's performing for a first iteration
<azonenberg>
Rise time 73 ps 10-90, 49 ps 20-80, without de-embedding the cable, and generally good looking waveform other than some high frequency ringing on the rising edge
<azonenberg>
It looks like there's probably some peaking around 5ish GHz but we'll see in more detail once i throw it on the VNA
<azonenberg>
So yes, there is indeed a big peak in the response in the 4-6 GHz range, but you can somewhat flatten that out
<azonenberg>
in post or with a filter or something
<azonenberg>
but the massive null at 1.8 GHz is another story
<d1b2>
<dannas> A bit off-topic. I believe @azonenberg created a board for practicing rework. Now I can't seem to find the schematics. I also seemed to recall him talking about eventually spinning another version of it. I thought it would be a fun thing to use for improving my lame soldering/rework skills.
<azonenberg>
dannas: correct. It's on my github as azonenberg/reworkctf but i do not recommend building it now
<azonenberg>
it uses an ancient spartan-3 fpga and i think one or two other EOL parts
<azonenberg>
and also there's bugs causing a few challenges to be a lot harder/easier than intended
<d1b2>
<dannas> @azonenberg Ok, in that case I better wait a couple of months then. Or are there other equiavelent boards you'd recommend?
<azonenberg>
I made it because nothing else existed
<azonenberg>
i would like to make another one but with the component shotrage it will be hard
<azonenberg>
in particular i want a low cost 1mm pitch bga component that i can hide some of the 500 point problems under
<azonenberg>
this basically means an entry level FPGA
<azonenberg>
and those are very hard to find right now
<d1b2>
<dannas> Ok, I'll watch the repo and see if there's any activity during the fall. The chip shortage situation is crazy. At work we payed 100x the price for STM32 chips. The electronic engineers spend most of their days finding replacements parts.
<azonenberg>
yeah I'm working on all of my probe stuff because resistors and obscure mixed signal stuff is kinda still possible to get
<azonenberg>
common MCUs? forget it
<azonenberg>
i'm sitting on about ten stm32f031s in 28 pin qfn from a project two years ago i forgot about and never did
<azonenberg>
now slowly trickling them out into boards
<azonenberg>
because it's the only mcu i can be sure of having available
<d1b2>
<dannas> @azonenberg I wonder if there's some niche of microcontrollers that there are plenty in supply of. Maybe PICs..
<d1b2>
<dannas> @azonenberg Thank you so much for all the work you've published on blogs, twitter and youtube. It's been really useful for me. I'll watch the reworkctf repo for activity but don't expect much to happen during the fall. And congratulations on the addition to your family (saw the notice on twitter).
<d1b2>
<Darius> "Hey our microcontroller sucks so much there's still stock"
<azonenberg>
Thanks
<azonenberg>
And lol, yes there are still some 8 bit pics in stock at digikey
<azonenberg>
There's actually STM32s too though
<azonenberg>
Just obscure ones
<azonenberg>
e.g. WLCSP packages are much more likely to be available than a common QFN or (heaven forbid) 0.8mm TQFP
<azonenberg>
the more obscure and high end a part is, the less likely it is to be gone
<azonenberg>
There are some exceptions though like FPGAs
<azonenberg>
There is a grand total of one xilinx 7 series fpga in stock at digikey
<azonenberg>
no, not one SKU
<azonenberg>
one physical chip
<azonenberg>
it's a $21K virtex-7 engineering sample
<azonenberg>
and i bet someone will be desperate enough to buy it soon
<d1b2>
<dannas> Someone somewhere is probably making the decision right now to use some very obscure part for short-time supply benefits.
<azonenberg>
Oh i expect that is happening all the time
<d1b2>
<dannas> And some other sw engineer is going to get notified in the last miute
<azonenberg>
people are respinning boards around whatever part they can get
<azonenberg>
There is a single kintex ultrascale on digikey too
<azonenberg>
Which is actually a part i might want, the ku025 in ffva1156, but i don't need it enough to drop $1K on it (especially when i won't be able to get all of the OTHER components i need to make a useful board for it)
<azonenberg>
and again, there is *one chip*
<azonenberg>
oh and status update on the probes
<azonenberg>
I'm now fairly sure that the issue is reflections in the path from the resistors to the amplifiers
<azonenberg>
because the 1.8 GHz null is almost perfectly centered for that
<azonenberg>
(I'm also going to try testing a differential mode signal vs having the other leg at ground)
bvernoux has joined #scopehal
someone-else has joined #scopehal
bvernoux has quit [Ping timeout: 240 seconds]
bvernoux has joined #scopehal
someone-else has quit [Quit: Connection closed]
CarlosEDP has quit [Quit: Bridge terminating on SIGTERM]
CarlosEDP has joined #scopehal
lain has quit [Ping timeout: 252 seconds]
lain has joined #scopehal
<d1b2>
<david.lenfesty> Yeah the reason I don't feel too bad about my lack of progress on the active probe interface is because nobody will be able to do anything with it for quite a while :P
someone-else has joined #scopehal
<bvernoux>
I have some news from Rigol Support about MSO5k Firmware update to speedup Data exchanged over Ethernet Gigabit
<bvernoux>
After the contact with the R&D. The acceleration improvement of the communication and the data capture of MSO5000 is not as simple as assumed. Therefore, we need more time. Its very difficult to give and share a new date, but I can give to you an update next 6-8 weeks.
<bvernoux>
It was the answer on 10 Aug 2021
<bvernoux>
So it seems we should have news in October 2021 if they not push it again ...
<bvernoux>
Does anyone here will be interested by a EMI/EMC LISN with Dual Channels(including 10dB Attenuator+Transient Limiter for SpectrumAnalyzer side) + DM CM Separator as I plan to build one for me but I will receive 5 PCBs
<bvernoux>
of each
<bvernoux>
It is a 5uH LISN for 60V / 5A max
<bvernoux>
I do not plan to build a 50uH LISN for >200V
<Stephie>
for 60V ac?
<Stephie>
that's an interesting voltage
<bvernoux>
Yes it support also AC in fact as there is 2 chan
<electronic_eel>
bvernoux: you plan to order the pcb design by user noy in the thread you linked?
<electronic_eel>
what kind of surface finish to you plan on using? ENIG or HASL?
<electronic_eel>
i'm a bit unsure what kind of surface finish is best. ENIG is best short term. but you want a good long term connection to the case. after a few years ENIG tends to turn black. but tinned also oxidizes
<electronic_eel>
i mean the connection between ground and the case, the outer ring around the circuit on the pcb
<bvernoux>
electronic_eel, Yes I have already ordered the PCB
<bvernoux>
I potentially plan to update the design with 4 Layers for an even better version
<bvernoux>
with perfect 50Ohms Impedance
<bvernoux>
as the components are pretty expensive
<bvernoux>
especially the big 1.1uH Wurth Inductances
<electronic_eel>
i just looked a bit at it and also saw that it was just 2 layer. while the circuit certainly doesn't need 4 layers, the rf performance would probably be improved by 4 layers
<electronic_eel>
how do you get 5µH with 5 pcs of 1.1 µH in series? wouldn't you use 1 µH instead?
<bvernoux>
It is a good question as it is 5.5uH ;)
<bvernoux>
It is why I want to check that with my VNA ;)
<electronic_eel>
the 1n4148 could be replaced by BAV99, probably you'd find something even better
<bvernoux>
yes BAV99 are very good
<electronic_eel>
i like BAV99 for cheap stuff, but there are some dedicated rf diodes available now, though more pricey
<bvernoux>
Potentially I plan to upgrade the actual version to 4 layers and add Improve like that
<bvernoux>
yes I cumulate BAV99 with dedicated ESD dioded ;)
<bvernoux>
as it add cumulative protection
<bvernoux>
the small 0402 ESD diode for RF are not good for big surge when BAV99 are good
<bvernoux>
so the both are a must have ;)
<electronic_eel>
if you have it on your bench and a vna connected, replacing the diodes and having a look at the performance will be easy
<bvernoux>
I'm the first to have introduced that on AirSpy
<bvernoux>
as there is both
<bvernoux>
and you shall have seen no anyone have burnt the Input ;)
<bvernoux>
as it resist to more than 30KV ESD on input
<electronic_eel>
there are 2 layers of surge protection on this board, before and after the 10db limiter as far as i can see. i'd keep BAV99 at the first layer and go to smaller rf diodes on the second
<bvernoux>
yes
<bvernoux>
Their protection seems pretty very strong I have not checked it carefully
<bvernoux>
Need to check that with simulation ;)
<bvernoux>
They have done lot of simulation in the EEVBlog thread
<electronic_eel>
you need strong protection on a lisn. you can easily burn your sa when using a lisn
<bvernoux>
yes
<bvernoux>
I have already a Transient Limiter in fact
<bvernoux>
From LeoBodnar it is very good
<bvernoux>
But that ones seems even better
<bvernoux>
Also very important point is to use 250V capacitor for the 100nF ones on DUT+/DUT-
<bvernoux>
For SMA connector I have replaced it by TE Connectivity 5-1814832-2
<electronic_eel>
why is the 250V part so important? wouldn't it be even better to use a NP0/C0G one?
<bvernoux>
it is 4 times cheaper and good enough as anyway a LISN is not intended to exceed 200MHz (or even >150MHz)
<bvernoux>
electronic_eel, because it is a power line up to 60V
<bvernoux>
so it requires a big capacitor with big voltage
<bvernoux>
which in addition allow to have big capacitance
<bvernoux>
for better filtering
<bvernoux>
NP0/C0G does not exist for such high voltage
<bvernoux>
I have selected C3216X7R2E104K160AA
<bvernoux>
It is good compromise quality/price and it is 250V
<bvernoux>
also 470nF Capacitor I have take some big ones ;)
<bvernoux>
100V
<bvernoux>
AVX 08051C474K4Z2A
<bvernoux>
and the very good things is they have integrated a GDT (Gas Discharge Tube)
<bvernoux>
so the transcient limiter is very good I think at end with all those protections
<electronic_eel>
i thought the GDT was primarily designed for use in the 50 µH variant for use with mains
<bvernoux>
the GDT is for both
<bvernoux>
it protect against big strike ;)
<electronic_eel>
i don't think a GDT will fire below 100 V
<electronic_eel>
and it is too slow for ESD
<bvernoux>
it is for big strike
<bvernoux>
with lot of current
<electronic_eel>
so i don't think it makes much sense for use in <60 V devices
<bvernoux>
it is 75V 5kA ;)
<electronic_eel>
you don't have any lightning strikes on your bench, don't you?
<bvernoux>
maybe in case of mistake ;)
<electronic_eel>
big mistake ;)
<bvernoux>
I think it help for ESD too
<bvernoux>
especially in USA/Canada where there is very big ESD
<electronic_eel>
how about using GRM31C5C2A104JA01L as capacitor. 100 V but NP0. I think it will perform much better than a X7R 250 V rated part
<bvernoux>
I see they have specified it shall be 250V
<bvernoux>
to be clarified yes
<bvernoux>
as the 5uH is not intended to exceed 60V / 5A anyway
<electronic_eel>
exactly
<bvernoux>
and in my case I do not design anything which exceed 60V ...
<bvernoux>
But yes I think 100V is good
<bvernoux>
anyway for other capacitor I have used 100V
<electronic_eel>
i sometimes do stress testing above 60V. but for regular use my devices usually are <=24V. maybe i do some 48V in the future
<bvernoux>
anyway 60V 5A is already very dangerous ;)
<electronic_eel>
60V gives nice long arcs when you switch off
<bvernoux>
The last point is to have a good software for the LISN
<bvernoux>
and I have not seen any open source software for it
<bvernoux>
TekBox have developped one which is expensive (>499Euros IIRC)
<bvernoux>
it is not so expensive when compared to R&S / KeySight LISN stuff ;)
<electronic_eel>
hmm, what do you need the software to do? can't just use the output from the sa and draw in the limit lines?
<bvernoux>
a dedicated software is even better
<bvernoux>
the one from TekBox or other vendor do the report too ;)
<electronic_eel>
the sa needs the option to use quasi peak detector of course
<bvernoux>
but yes just basic SA with a limit is enough
<bvernoux>
yes quasi peak detector is not available on all SA ...
<electronic_eel>
i don't plan to open a emc compliance test house soon. so i don't need to do fancy reports. it is just for precompliance during development
<bvernoux>
yes me too ;)
<bvernoux>
I have TekBox EMC Probe Set for EMC
<electronic_eel>
yeah, the near field probes. they are nice
<bvernoux>
it is very nice for the price even if some cheap EMC Probe set can be used too
<bvernoux>
I have the one with 40dB Amp
<electronic_eel>
my sa
<bvernoux>
so now the missing part is LISN ;)
<electronic_eel>
has an integrated preamp. so i don't need the external one
<bvernoux>
integrated preamp are often limited to 20dB here it is 40dB it is good to analyze very sensitive noise
<electronic_eel>
i'd like to have a current clamp for emi measurement. i built one myself based on some plans i found, using a big core with windings around, but the results are quite inconsistent
<electronic_eel>
also having a small tem cell would be nice
<bvernoux>
it is clearly very nice rf current monitoring probe
<bvernoux>
30KHz - 500MHz will be perfect ;)
<bvernoux>
but it cost 399USD
<electronic_eel>
emi problems below about 150 or 200 MHz are most often conducted through some cables, you'd get them better with such a current clamp than with an antenna
<electronic_eel>
also antennas in the range tend to be huge, so nothing you want in your small lab
<electronic_eel>
hmm, i guess the cal fixture is what was missing from my self built clamp
<bvernoux>
will be nice to have that in Open Source ;)
<bvernoux>
anyway it will be costly as it shall be built ...
<electronic_eel>
the measurements with my clamp varied wildly depending on the position of the cable in the clamp
<electronic_eel>
i don't know if this is normal and you are expected to use some holder or if the reason is some inconsistency in my winding or somehting like this
<bvernoux>
if you are interested to improve the LISN
<bvernoux>
I can publish my version which mainly include BOM with KiCost
<bvernoux>
and is converted to KiCad6
<bvernoux>
as the guys which have done the work on EEVBLOG forum have never published it on GitHub or similar
<bvernoux>
all that work lost in a forum dispatched in different threads ...
<electronic_eel>
there are some .zips linked in the thread. i think they contain the whole kicad project
<bvernoux>
yes I have based my work on latest zips
<bvernoux>
but it is a bit a mess ;)
<electronic_eel>
yes
<bvernoux>
especially the schematic include both version
<bvernoux>
i have removed everything related to 50uH in fact to clarify the LISN specialized for 5uH only
<bvernoux>
anyway the 50uH for >60V is too dangerous and I imagine not well validated
<bvernoux>
as it requires even more hw to validate that with high voltage
<electronic_eel>
you wrote you already ordered pcbs. i think shipping 2 layer pcbs from France to Germany is probably more expensive than me adding some 2layer pcbs to my next JLC order.
<electronic_eel>
but when you got yours and work on some improved 4 layer ones, i'd be interested to help and get some pcbs from you
<bvernoux>
3.5USD/unit and 1 LISN (with 2 chan) requires 10x
<bvernoux>
so about 35USD
<bvernoux>
The case is chezp Hammond 1590BBK
<bvernoux>
about 11USD
<electronic_eel>
yeah, still ok. often the case and connectors are more expensive than all the other parts
<bvernoux>
So far Total BOM (no including the PCB) is >86USD for the components
<electronic_eel>
i thought about building such a lisn before, but didn't finish it. i remember there were some other inductors proposed on some eevblog forum thread
<bvernoux>
So I think doing a 4 Layers with perfect 50Ohms Impedance with JLCPCB 4 Layers JLC7628 (FR-4 TG155) could be a must ;)
<bvernoux>
I have ordered a new test PCB using FR-4 TG155 + JLC7628
<electronic_eel>
i don't exactly remember which ones, but one should be able to find them
<bvernoux>
to check if the impedance is good
<electronic_eel>
the actual inductors are important, because how they cross couple will effect the results
<bvernoux>
with my new
<bvernoux>
4Layers_trl-board_v0_5
<electronic_eel>
so it would be interesting to compare the würth ones to the other ones
<bvernoux>
yes
<bvernoux>
The Wurth is very expensive
<electronic_eel>
the other ones were expensive too
<bvernoux>
The other TDK NL45323T-102J-PF is out of stock worldwide :(
<bvernoux>
and is something like 1USD/unit
<bvernoux>
it is because of chip shortage again ...
<bvernoux>
ha no it is available on Mouser
<bvernoux>
maybe I have mixed with an other part
<electronic_eel>
i guess. that doesn't look like the right part to me
<bvernoux>
<0.4Euro/unit so about 4 times cheaper vs Wurth
<bvernoux>
yes
<bvernoux>
ha yes I have made a mistake with the 1mH ;)
<electronic_eel>
1 mH
<electronic_eel>
that could be for the 10 db
<bvernoux>
yes
<bvernoux>
hmm I do not find the other replacement for 1.1uH
<bvernoux>
I'm pretty sure CoilCraft shall have something very good (probably better)
<bvernoux>
but even more expensive ;)
<bvernoux>
as CoilCraft are often crazy expensive
<electronic_eel>
got it: coilcraft SER2009-801MLD
<bvernoux>
about why 1.1uH instead of 1uH ;)
<bvernoux>
I think it is not very important as anyway Inductor have often a big tolerance like 10% or 5% in best case
<electronic_eel>
i think i have to read the thread again why they chose the 800nH instead of a 1µH one
<bvernoux>
yes
<electronic_eel>
but they are quite large, so i think i remember the idea was to do some 3d mounting tricks instead of simply soldering them down next to each other on the pcb
<bvernoux>
Does they provide important parameter for LISN use case ?
<bvernoux>
I imagine big current something like 10x5A
<bvernoux>
to be sure there is not saturation before to reach the max current ...
<bvernoux>
with Irms rise ...
<electronic_eel>
the capacitative and inductive coupling between the inductors is important. even if they are shielded they influence each other
<electronic_eel>
so it matters which ones are used and how they are mounted
<bvernoux>
yes it is why they have changed the PCB to remove 1 layer shield
<bvernoux>
but it is important to keep bottom layer GND shield
<bvernoux>
I remember of those constraints for NFC ;)
<bvernoux>
even if there is less voltage/current than in LISN which can reach up to 10A / 60V in fact even if here it is specified to 5A /60V
<bvernoux>
In my case I do not need 5A / 60V anyway ;à)
<bvernoux>
But who know if it can do more maybe in few years I will need to test a circuit with 5A / 40V or more ...
<electronic_eel>
if you do some motor driver or led lighting strip you'll quickly exceed the 5A
<bvernoux>
yes but so far I do not play with LED and Motor
<electronic_eel>
i'm annoyed by my kitchen lighting and how it is operated. maybe i install some led strips there and build a driver board for it
<electronic_eel>
a cap touch slider below the shelves would be nice to dim the led strips
<bvernoux>
For information I have asked the 2 Layers 1.6mm Specification to JLCPCB ;)
<bvernoux>
as there is clearly no any details on the PCB stackup for 2 layers ;)
<bvernoux>
I suspect the accuracy of layers is awful ;)
<electronic_eel>
i think they have several different factories for 2 layers and i think they use different stackups or processes
<bvernoux>
as their 2layers PCB are not intended to do 50 Ohms Impedance PCB ;)
<electronic_eel>
no
<bvernoux>
yes we will see what they answer
<bvernoux>
it seems to be not so critical for LISN but if we can have better => more accurate / more linear results it is a good move to use 4 layers with good 50 Ohms Impedance
<bvernoux>
Why not OSH Park 4 Layers as it is even better than FR4 ;)
<electronic_eel>
one thing i dislike about the layout of this model is that the distance between the input jacks is not 19mm. so you can't use bnc adapters and so on
<bvernoux>
but for LISN as the frequency does not exceed 150MHz it is not very important to have very good Er ...
<electronic_eel>
also having 4mm input connector with optional screw terminals would be nice
<electronic_eel>
either for big wires or for adding another filter on the psu side
<bvernoux>
they use Keystone 575-4
<bvernoux>
yes clearly there is improvement for that
<bvernoux>
to do
<bvernoux>
do you want I push my actual version ?
<bvernoux>
on github
<electronic_eel>
hirschmann 930136100 would be nice connectors for the input. but then you'd probably can't use the bottom mount pcb style anymore
<bvernoux>
I think the one they have is not the worst inside
<bvernoux>
but they shall respect 19mm space for BNC adaptor
<electronic_eel>
having the stuff on github would be nice for a start, not so chaotic as the forum thread
<bvernoux>
yes on the forum it is total chaos
<bvernoux>
on github at least we can add comment/Improvements step by step in a Wiki or in an Issue
<bvernoux>
let's push it ;)
<bvernoux>
I will add all credits to original author it is not easy to find their name
<bvernoux>
we can add that later I have not changed the KiCad description / author anyway
<bvernoux>
You are ok with my version which remove the 50uH version ?
<bvernoux>
or you was planning to build a 50uH version too ?
<electronic_eel>
making a 50µH version that is rated for mains is a completely different thing, much more involved
<electronic_eel>
i don't think it makes sense to try to combine these
<electronic_eel>
have a look at commercial 50µH lisns. they have huge inductors inside. not something you can buy off the shelf
<bvernoux>
yes clearly
<bvernoux>
it is why I have removed anything related to the 50uH
<bvernoux>
I will add a TODO in the schematic ;)
<bvernoux>
So let's start with different point to improve
<bvernoux>
So in TODO
<bvernoux>
TODO modifications/improvements of 5uH DC-LISN:
<bvernoux>
1) Add the Simulation in the GitHub repository and check everything (to be checked in EEVBLOG forum)
<bvernoux>
3-1) At least allow to have 19mm space between for BNC adapter