electronic_eel has quit [Ping timeout: 264 seconds]
electronic_eel has joined #scopehal
Degi_ has joined #scopehal
Degi has quit [Ping timeout: 265 seconds]
Degi_ is now known as Degi
futarisIRCcloud has joined #scopehal
_whitelogger has joined #scopehal
futarisIRCcloud has quit [Quit: Connection closed for inactivity]
_whitelogger has joined #scopehal
bvernoux1 has quit [Quit: Leaving]
bvernoux has joined #scopehal
<bvernoux>
I really hate this python stuff !!
<bvernoux>
I'm trying to let work nmigen on Ubuntu 18.04 LTS
<bvernoux>
and this "bastard" use an old Python 3.6.7 which is not compatible with nmigen stuff which seems to requires Python 3.8.x minimal like seen in RobertBaruch nmigen-tutorial
<bvernoux>
it is a real mess with Python3 to upgrade it correctly to 3.8.x ...
<bvernoux>
the same as it was a mess to switch from python2 to python3 ...
<miek>
i'm pretty sure it only requires 3.6+
<azonenberg>
i only use python if forced to kicking and screaming
<azonenberg>
i never properly learned it because every time i tried it was just unpleasant
<bvernoux>
it does not in fact see tutorial of Robert it requires Python 3.8
<azonenberg>
i'd rather write PHP than python and that says something :p
<bvernoux>
I understand it is very powerful but it have some many dependencies without any clear path to install all that mess
<bvernoux>
=> Install yosys, Symbiyosys, yices2, and z3
<bvernoux>
is a big mess ;)
<bvernoux>
nothing is clear
<bvernoux>
I have something working but it seems there is some option to set for ECP5 ...
<bvernoux>
so far the best tuto about nmigen is from RobertBaruch but it is far from perfection when doing it from scratch on Ubuntu 18.04 LTS (which is the recommended Linux distrib)
<bvernoux>
the 1st wrong things is the install does not work as Ubunut 18.04 LTS use Python 3.6.x instead of 3.8.x
<bvernoux>
so python 3.8.x shall be installed manually ...
<bvernoux>
to overload the default 3.6.x
<bvernoux>
ok Robert was recommending Ubuntu 20.04 LTS ;)
<bvernoux>
Which I suspect use by default Python 3.8.x that explain things ;)
<bvernoux>
on other tuto It was recommended Ubuntu 18.04 LTS but it seems it is obsolete as there is some dependencies with python 3.8 now with latest nmigen ...
<bvernoux>
so let's try again on an other VM with Ubuntu 20.04 lTS ;)
<bvernoux>
it is a long road to do a basic hello world on ECP5 ;)
<bvernoux>
ok I confirm Ubuntu 20.04 LTS after upgrade switch to Python3.8.x ;) which is not the case for Ubuntu 18.04 LTS which is still on Python 3.6.x
<bvernoux>
Ubuntu is clearly not coherent in the management on package too ...
<bvernoux>
why Ubuntu 18.04 LTS cannot use stable Python 3.8.x like 20.04 lTS !!!
<bvernoux>
I was keeping Ubuntu 18.04 LTS to have the choice to use Python2 or Python3
<bvernoux>
as there is still some old stuff requiring Python2 ...
_whitelogger has joined #scopehal
<bvernoux>
Also a fun things nmigen is available on 2 directories
<sorear>
it's an interpersonal mess between the owners of the respective organizations, not just a technical mess, and I don't feel qualified to summarize it
<bvernoux>
yes a total mess
<bvernoux>
mlabs seems to do not do anything on nmigen in fact
<bvernoux>
and forked it
<bvernoux>
they was sponsor of nmigen dev in fact
<bvernoux>
main issue is it is not clearly explained it is "like" a fork and active dev is done on https://github.com/nmigen/nmigen
<sorear>
whitequark left m-labs (the company) around the same time nmigen started, i don't know which caused which
<bvernoux>
ha ok
<azonenberg>
Yeah AFAIK wq was the main dev, left mlabs, and took nmigen with her
<bvernoux>
but mlabs never clarified that in their repo
<bvernoux>
so it is a real mess ;)
<azonenberg>
because mlabs still claims to be the official nmigen
<azonenberg>
and doesnt accept wq's version as the real thing
<bvernoux>
Imagine someone new to nmigen stuff which come on github and thinks mlabs is the main repo ;)
<bvernoux>
you are disappointed as nothing match especially for ECP56 ;)
<bvernoux>
ECP5
<bvernoux>
azonenberg, ha yes a real mess ;)
<azonenberg>
Also
<azonenberg>
First encapsulated AKL-PT2 is out of the mold
<azonenberg>
about to put it on the vna and characterize
<bvernoux>
ouhaou great
<bvernoux>
could you take a photo ?
<azonenberg>
I have one, just need to upload it
<azonenberg>
Gonna be a few
<bvernoux>
it is the version which shall go to 6GHz ?
<azonenberg>
also the clay i used to seal in a few spots is a pain to remove, i need to find a better process there
<bvernoux>
as I saw on tweeter you have tuned it to remove the 5.8GHz issue
<bvernoux>
so it shall be flat up to 6GHz
<azonenberg>
The version i have right now is the old pcb rev with 5.8 GHz null still. It's smooth rolloff, not flat. -3 dB is around 4 GHz right now due to losses in the long tail but with the notch fixed it will be usable to 6, just with >3 dB loss
<azonenberg>
if you can de-embed it's still usable
<bvernoux>
yes anyway it is already very nice
<azonenberg>
I also do not know how the silicone or the coverlay vs soldemrask will affect loss
<bvernoux>
so next version is intended to be more flat and with 6GHz -3dB ?
<azonenberg>
Next version is intended to have the 5.8 GHz null fixed and be whatever performance i can get. I'd be happy with 4 GHz from this version
<bvernoux>
ha yes it is very interesting to check effect of silicone ...
<azonenberg>
if i can get more, great
<azonenberg>
also need to work on the production version of the mold, fix some release issues in a few spots
<azonenberg>
maybe adjust the thickness of the mold so less friction on the sprues
<azonenberg>
make a ganged mold for multiple probes
<azonenberg>
maybe a few other tweaks for better overmolding
<azonenberg>
electrically with the notch fixed i am happy with performance
<azonenberg>
all remaining changes on the PT2 will be mechanical
<bvernoux>
yes anyway it is already very impressive
<bvernoux>
I never had such issue on Ubuntu 18.04 LTS
<bvernoux>
the 20.04LTS seems not very stable for upgrade ....
<azonenberg>
VNA data is in... looks like a few of the ripples in the response of the new probe shifted a bit but the overall shape is pretty smooth down to ~4 GHz -3dB bw
<azonenberg>
a few small dips at the early portion of the curve but this one also seems to have a cracked solder joint
<azonenberg>
i had to rework it a few times
<azonenberg>
It definitely needs the stiffener under the resistors
<azonenberg>
also probably a few changes to the mold around the SMA, it seems to be splitting
<sorear>
well you have _a_ answer
<azonenberg>
I think i'm probably going to make a second one just like this and do further testing on it. In any case though, there is not a huge increase in loss from the silicone
<azonenberg>
Which is the big question i had originally
<azonenberg>
The other thing i need to experiment with is the best way to get the silicone away from the resistors and castellations at the tip
<azonenberg>
the clay i was using is apparently not a good choice, it's hard to remove fully
<bvernoux>
woo very nice
<bvernoux>
in my case 4GHz BW is more than enough ;)
<bvernoux>
could be interesting to have more for test with VNA in fact
<_whitenotifier>
[starshipraider] azonenberg pushed 1 commit to master [+17/-17/±0] https://git.io/JL11O
<_whitenotifier>
[starshipraider] azonenberg 9f4265f - Renamed "resistive-probe" filenames to "AKL-PT2". Added new, snugger fit version of probe clip